AIML for Mechanical Engineers

Unit I
Introduction to AI and
ML in Mechanical
Engineering

Department of Mechanical Engineering, DYPCET, Kolhapur.

1
 AIML for Mechanical Engineers

Introduction to Artificial Intelligence and Machine Learning

➢ Artificial Intelligence (AI):
Artificial Intelligence, often shortened to AI, is a branch of computer science dedicated
to creating systems capable of performing tasks that typically require human intelligence.
These tasks encompass a broad spectrum, including understanding natural language,
recognizing patterns in data, making decisions based on complex information, and even
exhibiting behaviours associated with human-like reasoning.
AI systems are designed to simulate human cognitive processes such as learning,
problem-solving, and decision-making. Unlike traditional computer programs that follow
predefined instructions, AI systems can adapt and evolve based on experience and new
information. This ability to learn and improve over time is a key characteristic of artificial
intelligence.
In mechanical engineering, AI is employed in various applications, from designing
intelligent control systems for robotic arms to developing algorithms for predictive
maintenance in industrial machinery.

➢ Machine Learning (ML):

Machine Learning is a subset of AI that focuses on the development of algorithms
capable of learning from data and making predictions or decisions without being explicitly
programmed. The central idea behind machine learning is to enable computers to learn from
experience and improve their performance on a specific task over time.
In traditional programming, developers write code to instruct computers on how to
perform a task step by step. In contrast, in machine learning, algorithms are trained on large
datasets containing examples of input-output pairs. By analysing these examples, the algorithm
learns patterns and relationships in the data, which it can then use to make predictions or
decisions on new, unseen data.
Machine Learning algorithms can be categorized into several types, including
supervised learning, unsupervised learning, and reinforcement learning, each suited to different
types of tasks and data.
In mechanical engineering, Machine Learning is used for various purposes, such as
optimizing manufacturing processes, predicting equipment failures, and developing intelligent
control systems for autonomous vehicles.

➢ Relationship between AI and ML:

Artificial Intelligence and Machine Learning are closely related but distinct fields. AI
is a broader discipline that encompasses a wide range of techniques and approaches aimed at
creating intelligent systems. Machine Learning, on the other hand, is a specific subset of AI
focused on algorithms that enable computers to learn from data.
Machine Learning is a key component of Artificial Intelligence, providing the means
through which AI systems can learn and adapt to their environments. Many AI applications,
such as natural language processing, computer vision, and robotics, rely heavily on machine
learning techniques to achieve their goals.
In mechanical engineering, AI and ML work together to solve complex problems and
improve efficiency in various areas, including design optimization, predictive maintenance,
and autonomous systems. By leveraging the power of AI and ML, engineers can develop
innovative solutions to challenges in the field of mechanical engineering.

Department of Mechanical Engineering, DYPCET, Kolhapur.

2
 AIML for Mechanical Engineers

Fig. 1.1 AI and ML

1. AI Search Algorithms: When we want a computer to solve a problem, sometimes it needs

to look through a bunch of possible solutions. AI search algorithms help with that. They explore
different options systematically until they find the best solution. These algorithms are super
important for things like playing games (think of chess or Go), guiding robots to navigate
through complex environments, or planning routes for delivery services.

2. Logic: In AI, logic is like the language of reasoning. It helps computers understand
relationships between different pieces of information and draw conclusions. AI logic systems
use rules to represent knowledge and make decisions based on those rules. For example, they
can help in diagnosing diseases by analysing symptoms and medical data, or in planning routes
by considering traffic conditions and time constraints.

3. Knowledge Representation: Imagine you're trying to teach a computer everything you

know about a topic. Knowledge representation is about finding ways to store all that
information in a computer so that other AI programs can use it effectively. There are various
methods for representing knowledge, such as graphs, semantic networks, or logic-based
systems. Each method has its strengths and weaknesses, and choosing the right one depends
on the specific problem we're trying to solve.

4. Machine Learning: This is one of the most exciting fields in AI. Machine learning is all
about teaching computers to learn from data. Instead of programming explicit rules, we give
the computer examples and let it figure out patterns and relationships on its own. Machine
learning algorithms can be used for a wide range of tasks, such as recognizing handwriting,
predicting stock prices, or recommending movies on streaming platforms.

5. Natural Language Processing (NLP): Have you ever talked to a chatbot or asked Siri a
question? That's NLP in action. Natural language processing is about teaching computers to
understand and generate human language. NLP algorithms can analyze text to extract meaning,
translate between languages, or even generate human-like responses in conversations.

6. Vision: Computers can see, but not quite like humans do. Computer vision is the field of AI
that deals with teaching computers to interpret images and videos. This includes tasks like
object recognition (identifying what's in an image), image classification (sorting images into

Department of Mechanical Engineering, DYPCET, Kolhapur.

3
 AIML for Mechanical Engineers

categories), and image segmentation (dividing an image into meaningful parts). Computer
vision has applications in fields ranging from autonomous vehicles to medical imaging.

7. Robotics: Robotics combines engineering and AI to create machines that can sense, think,
and act in the physical world. Robots are becoming increasingly sophisticated, with sensors
that can detect objects and movements, algorithms that can make decisions based on sensory
input, and actuators that allow them to move and interact with their environment. AI plays a
crucial role in robotics by enabling robots to adapt to changing situations, learn from
experience, and perform complex tasks autonomously.

8. Expert Systems: Imagine having an expert in your pocket, ready to help you with any
problem you encounter. That's the idea behind expert systems. These are AI systems designed
to mimic the knowledge and reasoning skills of human experts in a specific domain. Expert
systems use a knowledge base of facts and rules, along with inference engines to draw
conclusions and make recommendations. They're used in various fields, including medicine
(for diagnosing diseases), finance (for making investment decisions), and engineering (for
troubleshooting technical problems).

Relation between Human Thoughts and Artificial Intelligence:

Fig. 1.2 Human and AI systems relationship

➢ Systems that think like humans:

Human thinking involves a complex interplay of cognitive processes such as logic,
reasoning, problem-solving, and learning. AI systems that aim to simulate human thought
processes often employ techniques such as pattern recognition, decision-making algorithms,
and neural networks.
While AI can replicate some aspects of human thinking, it typically lacks the nuanced
understanding, creativity, and intuition that humans possess. Human thought is influenced by
emotions, experiences, and social context, which are challenging for AI to fully emulate.

Department of Mechanical Engineering, DYPCET, Kolhapur.

4
 AIML for Mechanical Engineers

➢ Systems that act like humans:

Simulation of Human Thought Processes: AI systems designed to think like humans
aim to simulate cognitive processes such as logic, reasoning, problem-solving, and
learning. These systems often utilize techniques like machine learning, natural language
processing, and symbolic reasoning to mimic human-like thinking. Pattern Recognition and
Understanding: Just like humans, AI systems interpret and understand patterns in data,
enabling them to recognize objects, make predictions, and derive meaningful insights.
Emulating Creativity and Intuition: While AI can replicate certain aspects of human
creativity and intuition, it often lacks the depth and originality inherent in human thought.
However, advancements in generative AI are pushing the boundaries of creativity in
machine-generated content, such as art and music.
➢ Systems that think rationally:
Rational thinking involves making logical decisions based on available information and
reasoning processes. AI systems that think rationally use algorithms and rules to derive
conclusions and make predictions. Unlike humans, AI systems can process vast amounts
of data quickly and objectively, leading to efficient decision-making. However, they may
face challenges in dealing with uncertainty or ambiguity, which humans can navigate
through intuition and judgment.
➢ Systems that act rationally:
Physical Embodiment: AI systems that act like humans often involve robotics and
physical agents capable of performing tasks in the real world. These systems may exhibit
behaviours such as walking, talking, or manipulating objects, resembling human actions
and movements.
Natural Language Interaction: AI-powered chatbots and virtual assistants interact with
users using natural language processing techniques, providing responses and assistance
similar to human communication. These systems aim to emulate human conversational
patterns and understand user intentions.
➢ Relation between AI Thoughts and Human Thoughts:
AI thinking and behaviour are inspired by human cognition and behaviour but are
implemented using computational techniques and algorithms. While AI can simulate
certain aspects of human thought and behaviour, it often lacks the depth, complexity, and
context sensitivity inherent in human cognition.
Human thought and behaviour are shaped by biological, psychological, and social
factors, resulting in a rich tapestry of experiences, emotions, and interactions. AI, on the
other hand, operates within the constraints of algorithms and data, which may limit its
ability to fully replicate human-like thinking and behaviour.
Nonetheless, advances in AI research continue to bridge the gap between artificial and
human intelligence, leading to increasingly sophisticated systems that exhibit more
nuanced thinking and behaviour. However, achieving true human-like intelligence remains
an ongoing challenge in the field of AI.

Department of Mechanical Engineering, DYPCET, Kolhapur.

5
 AIML for Mechanical Engineers

System that acts like humans: Turing Test

Fig. 1.3 The Turing Test

“The art of creating machines that perform functions that require intelligence when
performed by people.” (Kurzweil)
“The study of how to make computers do things at which, at the moment, people are
better.” (Rich and Knight)

You enter a room which has a computer terminal. You have a fixed period of time to type
what you want into the terminal and study the replies. At the other end of the line is either a
human being or a computer system. If it is a computer system, and at the end of the period you
cannot reliably determine whether it is a system or a human, then the system is deemed to be
intelligent.

The Turing Test and AI Evaluation:

➢ Understanding the Turing Test:

The Turing Test, proposed by British mathematician and computer scientist Alan Turing
in 1950, is a benchmark for evaluating a machine's ability to exhibit intelligent behaviour
indistinguishable from that of a human. Named after Turing, the test was originally framed as
a thought experiment to explore the concept of machine intelligence.

➢ Key Components of the Turing Test:

1. Setup:
The Turing Test typically involves a text-based conversation between a human evaluator and
two participants: a human and a machine.
The evaluator interacts with both participants through a computer terminal, without knowing
which is the human and which is the machine.
2. Criterion:
The goal of the Turing Test is for the machine to convince the evaluator that it is human. If
the machine can successfully mimic human behaviour to the extent that the evaluator cannot
distinguish it from the human participant, it is said to have passed the test.
3. Nature of Interaction:
The conversation between the evaluator and the participants can cover a wide range of topics
and may involve questions, responses, and exchanges of information.
The machine's task is to generate responses that are linguistically and contextually
appropriate, demonstrating comprehension, reasoning, and coherence akin to human
communication.

Department of Mechanical Engineering, DYPCET, Kolhapur.

6
 AIML for Mechanical Engineers

➢ Implications of Passing the Turing Test:

1. Indication of Intelligence:
Passing the Turing Test has historically been regarded as a significant milestone in AI
research, suggesting that a machine possesses human-like intelligence.
The ability to engage in natural language conversation and exhibit behaviours indistinguishable
from humans implies a level of understanding and cognitive capability comparable to human
intelligence.
2. Challenges and Limitations:
Critics of the Turing Test argue that it may prioritize superficial imitation over genuine
understanding and intelligence. A machine could potentially pass the test by employing clever
strategies or pre-programmed responses without truly comprehending the content of the
conversation.
Additionally, the Turing Test focuses primarily on linguistic abilities and may not adequately
assess other facets of intelligence, such as creativity, emotional intelligence, or problem-
solving skills.

AI /ML Learning Life Cycle

Fig. 1.4 AI/ML Learning Life Cycle

1. Identifying the Problem: Setting the Course

In this first step, we're all about understanding the challenge ahead. We need to clearly
define the problem we want to solve and make sure it's something that machine learning can
tackle. Is it about sorting through emails to find spam? Or maybe predicting future sales trends?
By pinning down the problem, we set the direction for the entire journey.

2. Planning for Success: Strategizing and Resourcing

Once we've pinpointed the problem, it's time to map out our strategy. Here, we explore
different approaches to machine learning and figure out what resources we'll need. We're
talking about the technical stuff – like what algorithms and computing power we'll require. And

Department of Mechanical Engineering, DYPCET, Kolhapur.

7
 AIML for Mechanical Engineers

let's not forget about the people – assembling a team with the right mix of data science and
engineering skills is crucial to making our vision a reality.
3. Designing the MVP: Building a Prototype
Now it's time to roll up our sleeves and create a minimal viable product (MVP). Think
of it as a small-scale version of our final AI solution – something we can test with real data to
see how well it works. Building an MVP lets us iron out any kinks and tweak our approach
before we dive into full-scale development.

4. Developing and Refining: Iterating Towards Excellence

With insights from our MVP in hand, we move on to the next phase – developing and
fine-tuning our machine learning model. Here, we train the model with our data, constantly
checking its performance and making tweaks as needed. We're talking about things like
cleaning up the data, tweaking the features we're using, and selecting the best algorithms. Our
goal? To make sure our model is as accurate and reliable as possible, even when faced with
new data.

5. Deployment: Unleashing the Power

This is the moment we've been waiting for – putting our machine learning model to
work in the real world! We integrate it into our application or system, making it available to
the people who need it. And let's not forget about user training and support – we want to make
sure everyone knows how to use our new AI-powered tool.

6. Supporting and Maintaining: Ensuring Long-Term Value

But our journey doesn't end with deployment. In this final stage, we focus on keeping
our AI solution running smoothly over the long haul. That means keeping an eye on how it's
performing and making adjustments as needed. As new data comes in and the world changes,
we might need to update our model to keep it accurate and effective. It's all about making sure
our AI solution continues to deliver value over time.

Remember, the Lifecycle is Cyclical

Throughout this journey, we'll encounter decision points where we need to evaluate our
progress and decide whether to keep going or change course. If things aren't going as planned,
we might need to go back to an earlier stage and try a different approach. This cyclical nature
ensures that our AI development is flexible and adaptable, allowing us to evolve and improve
as we go.

By following these detailed stages and embracing the iterative nature of the lifecycle, we can
navigate the world of AI and machine learning with confidence, turning our ideas into impactful
solutions that make a real difference.

Department of Mechanical Engineering, DYPCET, Kolhapur.

8
 AIML for Mechanical Engineers

Evolution of AI/ ML

Fig. 1.5 Evolution of AI/ML

1950s: Early Excitement for Artificial Intelligence

The 1950s marked the dawn of artificial intelligence, igniting excitement and optimism
about the potential of creating intelligent machines. Pioneering work by scientists like Alan
Turing and John McCarthy laid the foundation for AI research and exploration.
1960s: Rise of Machine Learning
The 1960s saw the emergence of machine learning as a distinct field within AI.
Researchers began developing algorithms and techniques that enabled computers to learn from
data and improve their performance over time. This period laid the groundwork for many of
the machine learning approaches used today.
1970s: Period of Stagnation
Despite the initial enthusiasm for AI, the 1970s witnessed a slowdown in research and
progress. Challenges and limitations in computing power, data availability, and algorithmic
complexity led to a period of stagnation in the field of artificial intelligence.
1980s: Resurgence of Machine Learning
The 1980s saw a resurgence of interest and progress in machine learning. Advances in
computational capabilities and algorithmic techniques reignited enthusiasm for training
computers to learn from data and make predictions or decisions.
1990s: Continued Advancements in Machine Learning
Throughout the 1990s, machine learning continued to evolve and make significant
advancements. Researchers developed new algorithms and methodologies, leading to
improvements in areas such as pattern recognition, data mining, and predictive modelling.
2000s: Rise of Deep Learning
The 2000s marked a transformative period with the rise of deep learning. Breakthroughs
in neural network architectures, coupled with the availability of large datasets and powerful
computing resources, fueled an AI boom. Deep learning algorithms demonstrated remarkable
capabilities in tasks such as image recognition, speech recognition, and natural language
processing.
2010s: Deep Learning Dominance
The 2010s witnessed the continued dominance of deep learning as a driving force
behind AI advancements. Research breakthroughs in areas like convolutional neural networks
(CNNs) and recurrent neural networks (RNNs) propelled the field forward, leading to
significant improvements in AI performance across various domains.

Department of Mechanical Engineering, DYPCET, Kolhapur.

9
 AIML for Mechanical Engineers

AI Model Development

Fig. 1.6 AI Model Development

The diagram suggests that there is a way to train a machine learning model without
labelled data, It is a good starting point for understanding how machine learning models can
be trained without labelled data. Labelled data is data that has been manually labelled with the
correct output. For example, if you were training a machine learning model to recognize images
of cats, you would need to provide the model with a dataset of images that have been labelled
as "cat" or "not cat".

Training a machine learning model without labelled data is a challenging task, but it is
possible. AI Model Training Process for Identifying and Sorting Shapes. Training an AI model
to identify and sort shapes such as circles, triangles, and squares involves several steps. Here's
a detailed breakdown of the process:

1. Data Collection:
Gather a diverse dataset containing images of circles, triangles, and squares. Ensure
that the images cover various sizes, colors, orientations, and backgrounds to improve the
model's robustness. Label each image with its corresponding shape (e.g., circle, triangle,
square) to provide supervised training data for the model.
2. Data Preprocessing:
Resize all images to a consistent size to ensure uniformity in the input data.
Normalize pixel values to a common scale (e.g., between 0 and 1) to facilitate efficient training.
3. Model Architecture Selection:
Choose an appropriate neural network architecture for image classification tasks.
Convolutional Neural Networks (CNNs) are commonly used for this purpose due to their
ability to effectively capture spatial features in images.
Design the architecture with input layers, convolutional layers for feature extraction, pooling
layers for dimensionality reduction, and fully connected layers for classification.
4. Model Training:
Split the dataset into training, validation, and test sets. The training set is used to train
the model, the validation set is used to tune hyperparameters and monitor performance, and the
test set is used to evaluate the final model.

Department of Mechanical Engineering, DYPCET, Kolhapur.

10
 AIML for Mechanical Engineers

5. Evaluation and Validation:

Evaluate the trained model on the test set to assess its performance and generalization
ability. Calculate metrics such as accuracy, precision, recall, and F1-score to measure the
model's effectiveness in classifying shapes.
Visualize classification results and analyze any misclassifications to identify areas for
improvement.
6. Fine-Tuning and Optimization:
Fine-tune the model by adjusting hyperparameters, modifying the architecture, or
incorporating techniques such as data augmentation to further improve performance.
Optimize the model's computational efficiency by pruning unnecessary parameters or using
techniques like quantization for deployment on resource-constrained devices.
7. Deployment:
Deploy the trained model into production environments where it can be used to classify
shapes in real-time applications.
Integrate the model into an application or system, ensuring compatibility with input data
formats and providing user-friendly interfaces for interaction.
8. Continuous Monitoring and Maintenance:
Monitor the deployed model's performance in real-world scenarios and collect feedback
for ongoing improvement.
Periodically retrain the model with new data to adapt to changing patterns and ensure continued
accuracy and reliability.

Department of Mechanical Engineering, DYPCET, Kolhapur.

11
 AIML for Mechanical Engineers

Advantages of AIML

Fig. 1.7 Advantages of AI and ML

1. Boost Revenue:
AI and ML empower businesses to offer customized and personalized services to
customers, thereby enhancing customer satisfaction and loyalty, leading to increased revenue
streams.
2. Reduce Cost:
By implementing AI and ML technologies, mechanical engineering processes can be
optimized and automated, resulting in cost savings through improved efficiency, reduced
errors, and better resource utilization.
3. Reduce Error:
Automation processes enabled by AI and ML significantly reduce the occurrence of
errors in mechanical engineering tasks, leading to enhanced quality control and operational
reliability.
4. Unleash Potential:
AI and ML algorithms enable businesses to extract valuable insights from large
datasets, uncovering untapped opportunities for innovation, optimization, and growth in
mechanical engineering operations.
5. Enhanced Predictive Maintenance:
AI and ML algorithms analyze sensor data from mechanical systems to predict
equipment failures before they occur, allowing for proactive maintenance and minimizing
downtime.
6. Optimized Product Design:
AI and ML techniques optimize product design processes by analyzing historical data,
simulations, and feedback to iteratively improve product performance, reliability, and
efficiency.
7. Improved Supply Chain Management:
AI and ML algorithms optimize supply chain operations by forecasting demand,
optimizing inventory levels, and identifying opportunities for cost savings and efficiency
improvements.

Department of Mechanical Engineering, DYPCET, Kolhapur.

12
 AIML for Mechanical Engineers

Applications of AIML

Fig. 1.8 Applications of AI and ML

1. Automatic Language Translation:

AI-powered translation algorithms analyse text in one language and generate
corresponding translations in another language. This technology facilitates
communication across linguistic barriers and is widely used in applications such as online
content translation and multilingual customer support.
2. Speech Recognition:
Speech recognition systems convert spoken language into text, enabling hands-
free interaction with devices and applications. These systems power virtual assistants,
voice-controlled interfaces, and dictation software, improving accessibility and user
experience.
3. Image Recognition:
AI algorithms analyse visual data to identify objects, patterns, and features within
images. Image recognition technology is used in various applications, including facial
recognition, object detection, medical imaging analysis, and autonomous vehicles.
4. Product Recommendations:
AI-based recommendation systems analyse user behaviour and preferences to
suggest relevant products or content. These systems enhance user engagement, increase
sales, and personalize the customer experience in e-commerce platforms, streaming
services, and social media platforms.

Department of Mechanical Engineering, DYPCET, Kolhapur.

13
 AIML for Mechanical Engineers

5. Online Fraud Detection:

AI algorithms analyse patterns in user behaviour and transaction data to detect and
prevent fraudulent activities such as identity theft, payment fraud, and account takeover.
These systems protect businesses and consumers from financial losses and security
breaches.
6. Medical Diagnosis:
AI-driven diagnostic systems analyse medical data such as patient records, lab
results, and imaging scans to assist healthcare professionals in diagnosing diseases and
predicting patient outcomes. These systems improve accuracy, efficiency, and patient care
in medical diagnosis and treatment planning.
7. Stock Market Trading:
AI algorithms analyse financial data and market trends to make predictions about
stock prices and trading strategies. Automated trading systems use machine learning to
execute trades in real-time, optimizing investment decisions and portfolio management.
8. Traffic Prediction:
AI models analyse historical traffic data, weather conditions, and other factors to
predict traffic patterns and congestion levels. These predictions inform route planning,
traffic management, and transportation infrastructure planning, improving efficiency and
reducing commute times.
9. Virtual Personal Assistant:
Virtual assistants powered by AI/ML technology provide personalized assistance to
users in tasks such as scheduling appointments, setting reminders, and retrieving
information. These assistants, like Siri, Alexa, and Google Assistant, enhance productivity
and streamline daily tasks.
10. Email Spam and Malware Filtering:
AI algorithms analyse email content and user behaviour to identify and filter out
spam emails and malicious software. These systems protect users from phishing attacks,
malware infections, and other cybersecurity threats, safeguarding sensitive information and
data integrity.
11. Self-driving Cars:
AI algorithms analyze sensor data from cameras, radar, and lidar to perceive the
surrounding environment and make driving decisions autonomously. Self-driving cars use
machine learning to navigate roads, detect obstacles, and adapt to changing traffic
conditions, paving the way for safer and more efficient transportation systems.

Department of Mechanical Engineering, DYPCET, Kolhapur.

14
 AIML for Mechanical Engineers

Work-Flow of Artificial Intelligence

Fig. 1.9 Work-Flow of Artificial Intelligence

1. Data Acquisition:
This is where the journey begins. Data is gathered from various sources such as internal
databases, external data repositories, or even web scraping. This data could be anything from
customer information to sensor readings.
2. Data Analysis:
Once we have our data, it's time to dive in and explore. We analyse the data to uncover
patterns, trends, and insights. This involves using statistical methods, machine learning
techniques, or visualizations to understand the underlying structure of the data.
3. Data Cleansing and Preparation:
Before we can use our data for modelling, we need to clean it up. This involves
identifying and correcting errors, dealing with missing values, outliers, or inconsistencies.
Once our data is clean, we prepare it for use in a machine learning model by formatting it
appropriately.
4. Model Deployment:
With our data ready, we can now build our machine learning model. This model is
trained using the cleaned and prepared data, and once trained, it can be deployed into
production. In production, the model can make predictions on new, unseen data.
5. Optimization:
The work doesn't end once the model is deployed. We continually monitor its
performance and look for ways to improve it. This may involve tweaking the model's
hyperparameters, retraining it on new data, or even changing the model architecture altogether.

Overall, the process depicted in the image is cyclical, meaning it doesn't end once we've built
and deployed a model. Instead, it's an ongoing journey of acquiring, analysing, cleaning, and
using data to develop and refine machine learning models, ensuring they remain accurate and
effective over time.

Department of Mechanical Engineering, DYPCET, Kolhapur.

15
 AIML for Mechanical Engineers

Implementation of AI in Automation

Fig. 1.10 Implementation of AI in Automation

The Fig above gives a brief idea of how we can implement AI workflow in automation
system. The process consists of various steps which if implemented correctly enables us to
implement AI any advanced automation system. The steps are as follows:

1. Identify Data Source:

Begin by pinpointing where your data will originate. This could encompass a range of
sources such as customer databases, transaction records, social media interactions, or any other
data repositories relevant to your business operations.
2. Define Process:
Develop a comprehensive understanding of the workflow you intend to automate. This
involves mapping out each step of the current process, identifying decision points,
dependencies, and the roles of individuals involved in executing tasks.
3. Choose Technology:
Select the appropriate technology platform or workflow automation tool to integrate the
trained AI model. Consider factors such as ease of integration, scalability, compatibility with
existing systems, and support for the required functionalities
4. Train AI Model:
Utilize the data identified in the previous step to train a machine learning model. This
model learns from historical data patterns to make predictions or decisions that automate tasks
within the workflow. Training involves feeding the model with labelled data, allowing it to
recognize patterns and correlations.
5. Test & Improve Model:
Thoroughly evaluate the performance of the AI model through rigorous testing. This
includes running test scenarios using historical or simulated data to assess the accuracy and
reliability of model predictions. Based on testing outcomes, refine the model by adjusting
parameters, retraining with additional data, or fine-tuning algorithms.

Department of Mechanical Engineering, DYPCET, Kolhapur.

16
 AIML for Mechanical Engineers

6. Implement Model:
Once the AI model demonstrates satisfactory performance, proceed with its
implementation within the chosen workflow automation tool. Integrate the model seamlessly
into the workflow, configuring it to handle specific tasks and decision-making processes
efficiently.
7. Check Results:
Continuously monitor the AI-powered workflow to evaluate its effectiveness and
impact on business outcomes. Track key performance indicators (KPIs) to measure the success
of automation efforts and ensure alignment with organizational goals. Analyze model outputs,
identify any discrepancies, and refine the workflow as needed to optimize results.
8. Iterate and Refine:
Recognize that AI workflow automation is an iterative process. Regularly review and
refine both the AI model and the automated workflow to adapt to changing business
requirements, technological advancements, and evolving data patterns. Continuously seek
opportunities to enhance efficiency, accuracy, and scalability through ongoing iteration and
improvement.

By following this detailed workflow, organizations can effectively leverage AI and machine
learning technologies to automate complex business processes, drive operational efficiencies,
and achieve tangible business outcomes.

Department of Mechanical Engineering, DYPCET, Kolhapur.

17
 AIML for Mechanical Engineers

Working flow of Machine Learning

Fig. 1.11 Working Flow of Machine Learning.

1. Training Data: This is the data that the machine learning algorithm learns from. It is
important to have high-quality training data in order to train an accurate machine learning
model.
2. Machine Learning Algorithm: This is the code that learns from the training data and makes
predictions on new data. There are many different machine learning algorithms, and the best
algorithm for a particular task will depend on the nature of the data and the desired outcome.
3. Machine Learning Model: This is the output of the machine learning algorithm. It is a
representation of the knowledge that the algorithm has learned from the training data.
4. Predictions: These are the outputs that the machine learning model makes on new data. The
accuracy of the predictions will depend on the quality of the training data and the complexity
of the task.

The red line in the image represents the boundary between the training data and the machine
learning algorithm. The machine learning algorithm does not have access to the labels or target
variables of the training data during training. This is important because it helps to ensure that
the model is generalizing well and is not simply memorizing the training data.

The process of machine learning can be broken down into several steps:

1. Data Collection: This is the first step in the process, where data is collected from various
sources.
2. Data Preprocessing: This step involves cleaning and preparing the data for use in a
machine learning algorithm.
3. Model Selection: This step involves choosing a machine learning algorithm that is
appropriate for the task at hand.
4. Model Training: This step involves training the machine learning algorithm on the training
data.
5. Model Evaluation: This step involves evaluating the performance of the machine learning
model on a separate test dataset.
6. Model Deployment: This step involves deploying the machine learning model to
production.

Department of Mechanical Engineering, DYPCET, Kolhapur.

18
 AIML for Mechanical Engineers

Use Cases of AIML in Mechanical Engineering

Fig. 1.12 Man-Machine Interface (MMI): A Use Case of AI and ML

Here's a breakdown of the components in the MMI diagram:

User: This is the person interacting with the MMI, like a driver in a car or a technician in a
factory.
Knowledge Acquisition: It's about gathering information from the user, which helps control
the machine or provides feedback to the user.
Reasoning Machine: This part of the MMI processes the user's information and decides how
to control the machine.
Knowledge Process: It's about storing and retrieving information from the MMI's knowledge
base.
Case Base: It's a type of knowledge base holding specific examples of how problems were
solved before.
Rule Library: This knowledge base stores a set of rules used to make decisions.
Database: Another knowledge base storing factual information.
Learning System:
This allows the MMI to learn from experience, improving its performance over time.
In the diagram, arrows show how these components interact. For instance, the user interacts
with the knowledge acquisition part, sending info to the reasoning machine. The reasoning
machine then uses this to control the machine, possibly storing data in the knowledge process,
case base, or rule library. The learning system observes user-MMI interaction to enhance
performance.

Overall, MMIs are intricate systems with many parts. They work together to let users interact
with machines safely, efficiently, and easily. Whether you're driving a car or operating
machinery, MMIs make the experience smoother and more manageable.

Department of Mechanical Engineering, DYPCET, Kolhapur.

19
 AIML for Mechanical Engineers

Fig. 1.13 Different Modelling Techniques

The Fig. above is a diagram about different types of modelling techniques used in AI and
machine learning. Here's a detailed breakdown of the various techniques depicted in the
diagram:
• Data Rectification This process extracts meaningful features from raw data by eliminating
redundancies and noise. Data rectification employs recursive algorithms to clean the data.
• Predictive Modelling This technique involves training a machine learning model to forecast
future events or outcomes based on historical process data. The model can be used to predict
various aspects like sales figures, equipment failures, or customer churn.
• Process Optimization As the name suggests, this technique uses a trained AI model to
optimize a process for target parameters. The model can be continuously updated based on
real-time or previous process data to achieve the desired outcome.
• Fault Detection This AI-powered technique helps identify anomalies from the normal
process state. The model can pinpoint the root cause of these inconsistencies, allowing for
corrective actions to be taken.
• Process Control This technique involves training a machine learning model to establish and
regulate process parameters to maintain a desired target. The model can monitor and adapt
these parameters in real-time to optimize the process.
• Mechanistic Modelling This technique offers mechanistic insights into processes through
a combination of reverse engineering and employing AI models. It helps gain a deeper
understanding of how the various elements within a system function and interact with each
other.

Department of Mechanical Engineering, DYPCET, Kolhapur.

20
 AIML for Mechanical Engineers

Opportunities in Integrating AI in Mechanical Engineering

AIML is a language specifically designed for the creation of chatbots and virtual agents.
Following are the broad areas in which we have lots of opportunities in integration of AI and
ML in Mechanical Engineering.
• Design Optimization AI algorithms can analyze vast amounts of data to optimize designs.
This could involve optimizing for factors like material usage, weight reduction, or
structural integrity.
• Mechanical Systems Automation The design and optimization of mechanical systems and
parts may be automated using AI. This could involve tasks such as generating design
concepts or simulating the performance of different designs.
• Performance Simulation The performance of mechanical systems can also be simulated
and analysed using AI to forecast behaviour and suggest changes. This could be helpful in
predicting how a mechanical system will perform under different conditions or how it might
fail.
Overall, we can say that AI can be a powerful tool for mechanical engineers. By using AI for
design optimization, mechanical systems automation, and performance simulation, engineers
can create better products more efficiently.
It's important to note that while the concept might reference AIML, AI encompasses a
broader range of technologies than chatbot development. Machine learning, deep learning, and
natural language processing are some of the subfields of AI that are having a significant impact
on mechanical engineering.

Challenges and Considerations for AI integration in Mechanical Engineering

Integrating AI/ML in mechanical engineering poses several challenges, which can be classified
into four main types: mechanical data complexity, data quality, explainability, and ethical
challenges.

Mechanical Data Complexity:

Mechanical systems generate vast and complex data, including sensor readings,
performance metrics, and operational parameters. Handling this data requires sophisticated
algorithms capable of processing and analysing high-dimensional data efficiently. Additionally,
the diversity of data formats and structures in mechanical engineering poses challenges for
standardization and interoperability.
Data Quality:
Ensuring the quality and reliability of data used for training AI/ML models is critical
for achieving accurate results. Mechanical data may suffer from issues such as noise, missing
values, outliers, and measurement errors. Addressing these challenges requires robust data
preprocessing techniques, outlier detection algorithms, and data validation procedures to
enhance data quality before model training.
Explainability:
The black-box nature of many AI/ML algorithms presents challenges in understanding
and interpreting model predictions, especially in safety-critical applications within mechanical
engineering. Engineers need explanations and insights into how AI/ML models arrive at their
decisions to trust and validate their outputs. Developing interpretable and transparent AI/ML
models, such as explainable machine learning techniques, is essential for gaining user trust and
regulatory compliance.

Department of Mechanical Engineering, DYPCET, Kolhapur.

21
 AIML for Mechanical Engineers

Ethical Challenges:
Integrating AI/ML in mechanical engineering raises ethical considerations related to
privacy, bias, fairness, and accountability. Privacy concerns arise from the collection and use
of sensitive data, especially in applications involving personal information or proprietary
technology. Bias in AI/ML models can perpetuate unfair outcomes or discriminatory practices,
leading to social or legal ramifications. Ensuring fairness and equity in AI/ML algorithms
requires careful consideration of dataset biases and mitigation strategies. Moreover, ensuring
accountability and responsible AI deployment entails establishing clear guidelines, standards,
and governance frameworks to monitor and mitigate potential ethical risks throughout the AI
lifecycle.
Addressing these challenges requires interdisciplinary collaboration between
mechanical engineers, data scientists, ethicists, and policymakers to develop robust AI/ML
solutions that meet technical, ethical, and societal requirements in mechanical engineering
applications. Additionally, ongoing research and innovation are essential for advancing AI/ML
techniques tailored to the unique challenges and requirements of mechanical systems and
processes.

Department of Mechanical Engineering, DYPCET, Kolhapur.

22
 AIML for Mechanical Engineers

Unit II
AI Tools

Department of Mechanical Engineering, DYPCET, Kolhapur.

23
 AIML for Mechanical Engineers

AI Libraries AI Framework

AI Libraries and Frameworks:

AI libraries and frameworks are essential tools for developing and deploying artificial
intelligence applications. They provide pre-built functions and modules to streamline the
development process and offer a wide range of capabilities for machine learning and data
analysis.
In the realm of artificial intelligence (AI), libraries and frameworks play a pivotal role
in facilitating the development, deployment, and scaling of AI applications. These tools provide
a wide array of functionalities, ranging from implementing machine learning algorithms to
building complex neural networks. Let's delve into the key aspects of AI libraries and
frameworks:

Definition:
AI libraries and frameworks are collections of pre-written code modules, functions, and
utilities designed to simplify and expedite AI development tasks. They offer a range of tools
and functionalities tailored to various AI tasks, including data preprocessing, model training,
evaluation, and deployment.

Popular Libraries and Frameworks:

TensorFlow: Developed by Google, TensorFlow is one of the most widely used open-source
machine learning frameworks. It provides comprehensive support for building and training
deep neural networks, along with tools for deployment on various platforms.
PyTorch: PyTorch, maintained by Facebook's AI Research lab, is another popular deep
learning framework known for its dynamic computation graph and ease of use. It offers flexible
experimentation capabilities and extensive support for research prototyping.
Scikit-learn: Scikit-learn is a versatile machine learning library in Python, offering simple and
efficient tools for data analysis, preprocessing, and model training. It includes a wide range of
algorithms for classification, regression, clustering, and dimensionality reduction.
Keras: Keras is a high-level neural networks API written in Python, designed for fast
experimentation and prototyping of deep learning models. It offers a user-friendly interface and
seamless integration with TensorFlow and other backend engines.
MXNet: MXNet is an open-source deep learning framework developed by Apache, known for
its scalability and efficiency across distributed computing environments. It provides support
for both imperative and symbolic programming paradigms.

Department of Mechanical Engineering, DYPCET, Kolhapur.

24
 AIML for Mechanical Engineers

Caffe: Caffe is a deep learning framework developed by the Berkeley Vision and Learning
Centre (BVLC), optimized for speed and efficiency in image classification tasks. It is widely
used in computer vision applications.

Key Features of AI Libraries and Frameworks:

Flexibility: AI libraries and frameworks offer flexibility in terms of programming languages,

model architectures, and deployment options, allowing developers to choose the tools that best
suit their needs.
Scalability: Many frameworks are designed to scale seamlessly across multiple GPUs or
distributed computing environments, enabling the training of large-scale AI models on
extensive datasets.
Community Support: Popular libraries and frameworks benefit from active developer
communities, providing resources such as documentation, tutorials, and pre-trained models,
facilitating knowledge sharing and collaboration.
Integration: AI libraries and frameworks often integrate with other tools and platforms, such
as data visualization libraries, cloud computing services, and development environments,
streamlining the AI development workflow.

Application Areas:
AI libraries and frameworks find applications across various domains, including
computer vision, natural language processing, reinforcement learning, robotics, and healthcare.
They empower developers to build AI-driven solutions for diverse use cases, from image
recognition and language translation to autonomous driving and medical diagnosis.

In summary, AI libraries and frameworks serve as indispensable tools for developers

and researchers in harnessing the power of artificial intelligence to solve complex problems
and drive innovation across industries. Understanding their features, capabilities, and
application areas is essential for effectively leveraging AI technologies in real-world scenarios.

Now we will see Popular AI Tools in detail.

Department of Mechanical Engineering, DYPCET, Kolhapur.

25
 AIML for Mechanical Engineers

TensorFlow: A Comprehensive Overview

• Introduction: TensorFlow is an open-source machine learning framework developed by
Google Brain for building and training deep learning models. It was initially released in
2015 and has since become one of the most widely used and influential tools in the field of
artificial intelligence.
• Development: TensorFlow was developed by the Google Brain team to address the need
for a scalable and flexible framework for conducting research in machine learning and
deploying production-ready AI applications. It was built upon the lessons learned from
previous internal Google projects like DistBelief and was designed to be fast, efficient, and
easy to use.

Features:
• Flexibility: TensorFlow offers a flexible architecture that supports both high-level APIs for
easy model development and low-level APIs for fine-grained control over model
components.
• Scalability: It can efficiently scale from running on a single CPU to distributed computing
environments with multiple GPUs or even across clusters of machines.
• Comprehensive Tooling: TensorFlow provides a comprehensive suite of tools for data
preprocessing, model training, evaluation, and deployment, including TensorBoard for
visualization and TensorFlow Serving for serving trained models in production
environments.
• Extensibility: It supports customization and extensibility through its rich ecosystem of
libraries, extensions, and integration with other popular machine learning frameworks.
• Support for Various Platforms: TensorFlow supports deployment across a wide range of
platforms, including desktops, servers, mobile devices, and cloud environments, making it
suitable for both research and production use cases.

Applications:
• Computer Vision: Image classification, object detection, image segmentation.
• Natural Language Processing (NLP): Text classification, sentiment analysis, language
translation.
• Speech Recognition: Voice recognition, speech synthesis, speaker identification.
• Reinforcement Learning: Game playing, robotics, autonomous systems.
• Healthcare: Medical image analysis, disease diagnosis, drug discovery.

Advantages:
• Performance: TensorFlow offers high performance and efficiency, especially when
running on GPUs or distributed computing environments.
• Flexibility: Its flexible architecture supports experimentation and customization, allowing
researchers and developers to implement a wide range of machine learning algorithms and
models.
• Scalability: TensorFlow can scale seamlessly from training small models on a single device
to training large-scale models on distributed systems.
• Community Support: TensorFlow benefits from a large and active community of
developers, researchers, and enthusiasts who contribute to its ongoing development,
provide support, and share resources such as tutorials, libraries, and pre-trained models.

Department of Mechanical Engineering, DYPCET, Kolhapur.

26
 AIML for Mechanical Engineers

TensorFlow Workflow

The image above shows a high-level overview of a TensorFlow workflow, specifically focused
on building and deploying machine learning models. Here's a breakdown of the steps involved:
• Input Function (tf.data): This stage involves preparing and feeding data into the
TensorFlow pipeline. It ensures the data is formatted appropriately for the machine learning
model.
• Model (model_fn): This represents the core machine learning model you're building using
TensorFlow. The model definition outlines the layers, architecture, and hyperparameters
that will be used for training.
• Train: This step refers to the training process where the machine learning model is trained
on the provided data using the defined model architecture.
• Evaluate: After training the model, it's crucial to evaluate its performance on a separate
dataset to assess its effectiveness and identify areas for improvement.
• Checkpoint: This refers to periodically saving the model's state during training. These
checkpoints can be used to resume training later or roll back to a previous state if necessary.
• Predict: Once trained and evaluated, the model can be used to generate predictions on new,
unseen data.
• TensorFlow Serving: This refers to deploying the trained model for real-world usage.
TensorFlow Serving allows you to serve the model through APIs or web applications,
enabling it to make predictions on data received from external sources.
• TensorFlow Lite: This step represents an optional pathway for deploying the model on mobile
and embedded devices. TensorFlow Lite is a lightweight framework that converts the
TensorFlow model into a smaller, more efficient format suitable for resource-constrained
environments.
• Data Files: These represent the raw data used to train the machine learning model. The file
format can vary depending on the data type (images, text, etc.).

Overall, the image depicts a simplified overview of the TensorFlow machine learning workflow,
encompassing data preparation, model building, training, evaluation, deployment, and potential
conversion for mobile/embedded devices.

Department of Mechanical Engineering, DYPCET, Kolhapur.

27
 AIML for Mechanical Engineers

Keras: A Comprehensive Overview

Introduction:
Keras is an open-source neural networks library written in Python. It was developed with a
focus on enabling fast experimentation and prototyping of deep learning models. Keras was
designed to be user-friendly, modular, and extensible, making it a popular choice among
researchers and practitioners in the field of artificial intelligence.

Development:
Keras was originally developed by François Chollet, a Google engineer, as part of the research
project ONEIROS (Open-ended Neuro-Electronic Intelligent Robot Operating System). It was
released in March 2015 and has since gained widespread adoption due to its simplicity and ease
of use.

Features:
User-Friendly Interface: Keras offers a simple and intuitive API that allows users to define
and train neural networks with minimal code. It abstracts away the complexities of deep
learning frameworks, making it accessible to beginners and experts alike.
Modularity: Keras follows a modular design philosophy, allowing users to construct neural
network models by assembling pre-built layers and modules. This modular approach enables
rapid experimentation and facilitates the creation of complex network architectures.
Flexibility: It provides support for both convolutional neural networks (CNNs) and recurrent
neural networks (RNNs), as well as a wide range of activation functions, optimizers, and loss
functions. Additionally, Keras can seamlessly integrate with other deep learning libraries such
as TensorFlow and Theano.
Extensibility: Keras allows users to easily extend its functionality by creating custom layers,
loss functions, and callbacks. This extensibility enables the implementation of advanced deep
learning techniques and architectures.
Compatibility: Keras is compatible with multiple backends, including TensorFlow, Theano,
and Microsoft Cognitive Toolkit (CNTK). This compatibility ensures that users can leverage
the performance benefits of different backend engines while retaining the same high-level API.

Applications:
Image Classification: Identifying objects in images, such as cats, dogs, or cars.
Natural Language Processing (NLP): Text classification, sentiment analysis, named entity
recognition.
Sequence Modelling: Time series forecasting, speech recognition, language translation.
Generative Modelling: Generating images, text, or music using generative adversarial
networks (GANs) or variational autoencoders (VAEs).
Advantages:
Simplicity: Keras offers a straightforward and easy-to-understand interface, making it suitable
for beginners and researchers without deep learning expertise.
Flexibility: Its modular design and extensible architecture enable users to create custom neural
network architectures tailored to their specific requirements.

Department of Mechanical Engineering, DYPCET, Kolhapur.

28
 AIML for Mechanical Engineers

Integration: Keras seamlessly integrates with popular deep learning frameworks such as
TensorFlow, allowing users to leverage the performance benefits of these backends while using
the same high-level API.
Community Support: Keras benefits from a large and active community of developers,
researchers, and enthusiasts who contribute to its ongoing development, provide support, and
share resources such as tutorials, libraries, and pre-trained models.

Machine Learning vs Deep Learning through Application

The image above is related to a high-level comparison between traditional machine learning
and deep learning approaches. Here's a breakdown of the key elements in the image:

Traditional Machine Learning

• Features - Traditional machine learning models typically require domain experts to hand-
craft features from raw data. This feature engineering process involves selecting and
extracting relevant characteristics that best represent the data for the model to learn from.
• Model - The model in traditional machine learning is often a simpler algorithm, such as
linear regression or decision trees. These models can be effective for specific tasks, but
they may not be as powerful as deep learning models for complex problems.

Deep Learning
• Raw Data - Deep learning models can work directly with raw data, such as images, text,
or sensor readings. This eliminates the need for manual feature engineering, as the deep
learning model can automatically learn the relevant features from the data itself.
• Hidden Layers - Deep learning models contain multiple hidden layers between the input
and output layers. These hidden layers allow the model to learn complex relationships
between the features in the data. The more hidden layers a model has, the deeper it is said
to be.
• Deep Neural Network - Deep learning models are essentially deep neural networks, which
are inspired by the structure and function of the human brain. These neural networks consist
of interconnected nodes (artificial neurons) that process information layer by layer.

Department of Mechanical Engineering, DYPCET, Kolhapur.

29
 AIML for Mechanical Engineers

Key Differences
The image highlights the key differences between traditional machine learning and deep
learning:
• Feature Engineering - Traditional machine learning requires manual feature engineering,
while deep learning can learn features automatically.
• Model Complexity - Deep learning models can be more complex than traditional machine
learning models, with multiple hidden layers and a deeper architecture.
• Data Requirements - Deep learning models often require more data to train effectively
compared to traditional machine learning models.

Overall, the image depicts the advantages of deep learning over traditional machine learning
for complex tasks. Deep learning's ability to learn features automatically and its deeper
architecture allows it to capture intricate patterns in data, making it a powerful tool for various
applications.

Sequential Model and Functional Model in Keras

In Keras, a popular deep learning library, there are two primary ways to build neural network
architectures: the Sequential model and the Functional API. Let's explore each approach:
1. Sequential Model:
The Sequential model is the simplest and most commonly used way to build neural
networks in Keras. It allows you to create models layer-by-layer in a linear fashion, where each
layer has exactly one input tensor and one output tensor. This model is suitable for building
straightforward architectures, such as feedforward neural networks or simple convolutional
neural networks (CNNs).
You can create a Sequential model and add layers to it using the Sequential class and
the add() method, like this: The Sequential model is easy to understand and implement, making
it ideal for beginners and quick prototyping.
2. Functional Model:
The Functional API provides a more flexible and powerful way to build neural network
architectures compared to the Sequential model. It allows you to create models with complex
architectures, including multiple inputs, multiple outputs, shared layers, and branching
networks.
With the Functional API, you can define multiple input tensors and output tensors and explicitly
connect layers together to create a directed acyclic graph (DAG) of layers.
This model is suitable for building advanced architectures, such as siamese networks,
multi-input/multi-output networks, and models with residual connections. You can create
models using the Functional API by instantiating keras. Model objects and connecting layers
using the functional Model class and the keras. layers module.

Department of Mechanical Engineering, DYPCET, Kolhapur.

30
 AIML for Mechanical Engineers

• Keras is an API that was made to be easy to learn for people. Keras was made to be simple.
• Prototyping time in Keras is less. This means that your ideas can be implemented and
deployed in a shorter time.
• The research community for Keras is vast and highly developed.

Advantages of Keras:

Ease of Use and Enjoyability

• Keras is known for its user-friendly and concise syntax, making it easier for developers
to build and experiment with deep learning models compared to frameworks like
TensorFlow that require more low-level coding [2].
Faster Development
• Keras provides high-level building blocks like pre-trained models and modular
components that simplify and accelerate deep learning model development [2]. This
can save developers time and effort compared to building models from scratch.
Strong foundation for Complex Models
• Despite its ease of use, Keras allows you to create complex deep learning architectures
by stacking multiple layers and exploring various neural network configurations [2].
Support for Multiple Platforms and Backends
• Keras acts as a high-level interface that runs on top of different deep learning
frameworks like TensorFlow, PlaidML, or Theano [2]. This flexibility allows
developers to choose the backend that best suits their needs and hardware environment.
Large Open-Source Community
• Keras benefits from a vast open-source community that actively contributes to its
development and provides a wealth of resources, tutorials, and pre-trained models [2].
This can significantly aid developers in learning and working with Keras.
Applications:
➢ Keras is used for creating deep models which can be productized on smartphones.
• Keras is also used for distributed training of deep learning models.
• Keras is used by companies such as Netflix, Yelp, Uber, etc
• Keras is optimized for speed and efficiency. It allows us to train large and complex models
in a reasonable amount of time.

Department of Mechanical Engineering, DYPCET, Kolhapur.

31
 AIML for Mechanical Engineers

• MATLAB provided all the necessary tools to successfully apply AI in imaging and
radiomics.
• MATLAB was used for tasks such as data preprocessing, image segmentation.
• MATLAB is a programming platform designed specifically for engineers and scientists to
analyze and design systems and products.
• The heart of MATLAB is the MATLAB language, a matrix-based language expression of
computational mathematics.
• Data Analysis and Visualization: MATLAB provides powerful tools for data analysis,
manipulation, and visualization. It offers functions for data cleaning, filtering, statistics,
plotting, and creating interactive visualizations.
• Algorithm Development: MATLAB is widely used for developing and implementing
algorithms.

The image above is a diagram depicting the process of model tuning in MATLAB. Model
tuning is an iterative process that involves adjusting the hyperparameters of a machine learning
model to improve its performance on a given task. Here's a detailed breakdown of the steps
outlined in the diagram:
Select Features
• This initial step involves choosing the features or input variables that will be used to train
the machine learning model. Selecting a relevant set of features is crucial for achieving
optimal model performance.
Preprocessed Data
• This refers to the data that has been prepared for use in the machine learning model.
Preprocessing may involve steps such as cleaning the data, handling missing values, and
scaling the features to a common range.
Extract Features
• In some cases, you might extract additional features from the raw data to improve the
model's ability to learn complex patterns. Feature extraction techniques can involve
mathematical transformations or domain-specific knowledge.

Department of Mechanical Engineering, DYPCET, Kolhapur.

32
 AIML for Mechanical Engineers

Model Selection & Training

• Here, you choose a machine learning algorithm suitable for the task at hand and train it on
the pre-processed data and selected features. The training process involves fitting the
model to the data to learn the underlying relationships between the features and the target
variable.
Optimize Hyperparameters
• This is the core aspect of model tuning. Hyperparameters are settings that control the
learning process of a machine learning algorithm. Examples include the number of hidden
layers in a neural network or the learning rate used for gradient descent optimization.
Tuning involves adjusting these hyperparameters to optimize the model's performance on
a validation dataset.
Assess Performance
• It's essential to evaluate the performance of the tuned model on a separate validation
dataset to assess its generalizability and avoid overfitting. Overfitting refers to a scenario
where the model performs well on the training data but poorly on unseen data.
Deploy & Integrate
• Once a well-performing and tuned model is obtained, you can deploy it for real-world use.
This may involve integrating the model into a production environment or web application
to make predictions on new data.

Overall, the image highlights the importance of model tuning in achieving optimal performance
from machine learning models in MATLAB. By following a structured approach that involves
selecting features, preprocessing data, training the model, optimizing hyperparameters, and
evaluating performance, data scientists can leverage MATLAB's capabilities to build robust
and effective machine learning models. MATLAB provides powerful tools for data analysis,
manipulation, and visualization. It offers functions for data cleaning, filtering, statistics,
plotting, and creating interactive visualizations.

AI Role in Mechanical Engnieering

With AI becoming more integrated into mechanical engineering, engineers now have
AI-enhanced roles. They work with AI algorithms and tools to optimize designs, predict
maintenance needs, and improve overall system performance.
Materials play a crucial role in mechanical engineering, and AI is transforming the way
engineers approach material design and selection. Machine learning in mechanical engineering
algorithms can analyze vast databases of material properties, helping engineers identify novel
materials with specific characteristics.
Simulation and Analysis: AI can assist in simulating and analyzing complex engineering
systems. Machine learning algorithms can learn from historical data to predict system
behavior, identify potential failures or performance issues, and recommend design
modifications

Department of Mechanical Engineering, DYPCET, Kolhapur.

33
 AIML for Mechanical Engineers

AI in Predictive Maintenance
By using machine learning (ML) algorithms to underpin larger AI frameworks,
companies can collect historic and current data to anticipate failures before they happen
and take action to reduce the risk. AI systems can analyze historical and real-time
equipment, inventory, and purchasing data to help maintenance teams maintain optimal
inventory levels. These systems can identify patterns in your parts usage and purchase
history to recommend when you should restock parts
In manufacturing, predictive maintenance using AI helps minimize unplanned
downtime by analyzing sensor data from machines and equipment. ML algorithms can detect
anomalies and patterns in data, predicting when maintenance is needed.

Machine Learning Model

The image above is block diagram of a simple machine learning model. The block diagram
shows how a simple machine learning model works step-by-step. Here's a breakdown of the
different stages:
• Training Data: This is the data that the machine learning algorithm learns from. It is
important to have high-quality training data in order to train an accurate machine learning
model. In the image, the training data goes into a rectangular box labelled "Data".
• Data Preprocessing: This step involves cleaning and preparing the data for use in a machine
learning algorithm. Common preprocessing tasks include handling missing values,
normalization, and feature scaling. The image depicts this stage with a rectangular box
labelled "Preprocess".
• Feature Engineering: This step involves creating new features from the raw data that might
be more informative for the machine learning model. This is an optional step, but it can be
important for improving the performance of the model. The image does not explicitly show
this stage, but it can be incorporated between "Preprocess" and "Model" blocks.
• Model: This is the code that learns from the training data and makes predictions on new
data. There are many different machine learning algorithms, and the best algorithm for a
particular task will depend on the nature of the data and the desired outcome. The image
represents the model as a rectangular box labelled "Model".
• Learning Algorithm: This is the algorithm that is used to train the machine learning model.
The learning algorithm takes the training data as input and outputs a model that can be used
to make predictions on new data. The specific learning algorithm used will depend on the

Department of Mechanical Engineering, DYPCET, Kolhapur.

34
 AIML for Mechanical Engineers

type of machine learning model being used. The image does not differentiate between
"Model" and "Learning Algorithm".
• Predictions: These are the outputs that the machine learning model makes on new data. The
accuracy of the predictions will depend on the quality of the training data and the complexity
of the task. The image shows the predictions coming out of a rectangular box labelled
"Predictions".

The red line in the image represents the boundary between the training data and the machine
learning model. The machine learning algorithm does not have access to the labels or target
variables of the training data during training. This is important because it helps to ensure that
the model is generalizing well and is not simply memorizing the training data.

The process of machine learning can be broken down into several steps:

1. Data Collection: This is the first step in the process, where data is collected from various
sources.
2. Data Preprocessing: This step involves cleaning and preparing the data for use in a machine
learning algorithm.
3. Model Selection: This step involves choosing a machine learning algorithm that is
appropriate for the task at hand.
4. Model Training: This step involves training the machine learning algorithm on the training
data.
5. Model Evaluation: This step involves evaluating the performance of the machine learning
model on a separate test dataset.
6. Model Deployment: This step involves deploying the machine learning model to
production.

Machine learning is a powerful tool that can be used to solve a wide variety of problems.
However, it is important to remember that machine learning is not a magic bullet. The success
of a machine learning project depends on a number of factors, including the quality of the data,
the choice of machine learning algorithm, and the expertise of the data scientist.

Department of Mechanical Engineering, DYPCET, Kolhapur.

35
 AIML for Mechanical Engineers

The image you sent appears to be a diagram of a data science project lifecycle, including the
main stages involved in bringing a data science project from ideation to production. Here's a
breakdown of the stages depicted in the diagram:
1. Business Understanding: This initial stage involves understanding the business problem
or opportunity that the data science project aims to address. It's crucial to clearly define the
goals and objectives of the project to ensure it aligns with the overall business strategy.
2. Data Acquisition: Here, data scientists gather the relevant data required to build the
machine learning model. This data may come from various sources, including internal
databases, external data providers, or web scraping. Data quality is essential at this stage,
as it significantly impacts the performance of the model.
3. Data Understanding: Once the data is collected, data scientists explore and analyze it to
gain insights into its properties, distribution, and potential challenges. This stage involves
data cleaning, handling missing values, and identifying anomalies or outliers.
4. Data Preparation: In this stage, the data is preprocessed to prepare it for use in a machine
learning model. Preprocessing steps may involve scaling numerical features, encoding
categorical features, and feature engineering to create new features that might be more
informative for the model.
5. Modelling: This stage involves selecting a suitable machine learning algorithm and
training the model on the prepared data. The choice of algorithm depends on the nature of
the problem and the type of predictions you want to make. Some common machine learning
algorithms include linear regression, decision trees, random forests, and support vector
machines.
6. Model Evaluation: After training the model, it's crucial to evaluate its performance on a
separate test dataset to assess itsgeneralizability and avoid overfitting. Overfitting refers to
a scenario where the model performs well on the training data but poorly on unseen data.
Common metrics used for model evaluation include accuracy, precision, recall, and F1-
score.

Department of Mechanical Engineering, DYPCET, Kolhapur.

36
 AIML for Mechanical Engineers

7. Deployment: If the model performs well on the test dataset, it can be deployed to
production. This may involve integrating the model into a web application, API, or data
pipeline to make predictions on new, real-world data.
8. Monitoring: Once deployed, it's essential to monitor the model's performance over time to
ensure it continues to make accurate predictions. Monitoring may involve tracking model
metrics and retraining the model with new data if its performance degrades.
The cyclical nature of the arrows in the diagram emphasizes that data science projects are
iterative. As you learn more from the data and the deployed model, you can revisit earlier stages
to refine your approach or improve the model's performance.
Overall, the image provides a high-level overview of the data science project lifecycle,
highlighting the key stages involved in taking a project from conception to real-world
implementation.

Department of Mechanical Engineering, DYPCET, Kolhapur.

37
 AIML for Mechanical Engineers

Unit III
Machine Learning

Department of Mechanical Engineering, DYPCET, Kolhapur.

38
 AIML for Mechanical Engineers

Linear Regression
Introduction:
Linear regression is a fundamental supervised learning technique used for modelling the
relationship between a dependent variable (target) and one or more independent variables
(features). It assumes a linear relationship between the input variables and the output variable,
making it one of the simplest and most widely used regression methods.
Linear regression analysis is used to predict the value of a variable based on the value
of another variable. The variable you want to predict is called the dependent variable. The
variable you are using to predict the other variable's value is called the independent variable.

Linear regression shows the linear relationship between the independent(predictor) variable i.e.
X-axis and the dependent(output) variable i.e. Y-axis, called linear regression.
Simple linear regression is used to estimate the relationship between two quantitative variables

Key Concepts:

Linear Model: In linear regression, the relationship between the independent variables

Objective: The goal of linear regression is to find the best-fitting line (or hyperplane in higher
dimensions) that minimizes the difference between the predicted values and the actual values
of the target variable.

Department of Mechanical Engineering, DYPCET, Kolhapur.

39
 AIML for Mechanical Engineers

Types of Linear Regression:

Simple Linear Regression: In simple linear regression, there is only one independent variable
𝑥.

Linear Regression Formulae

Multiple Linear Regression: In multiple linear regression, there are multiple independent
variables that are used to predict the dependent variable 𝑦.

The relationship between the independent variables and the dependent variable is represented
by a hyperplane in a multidimensional space.

Assumptions of Linear Regression:

Linearity: The relationship between the independent and dependent variables is assumed to be
linear.
Independence: The observations are assumed to be independent of each other.
Homoscedasticity: The variance of the error terms is constant across all levels of the
independent variables.
Normality: The error terms are assumed to be normally distributed.

Applications of Simple Linear Regression:

Economics: Predicting sales, forecasting stock prices, analysing demand-supply relationships.
Finance: Predicting asset returns, modelling risk factors, evaluating investment strategies.
Healthcare: Predicting patient outcomes, analysing the relationship between risk factors and
diseases.
Social Sciences: Analysing survey data, studying the impact of interventions on behaviour.

Department of Mechanical Engineering, DYPCET, Kolhapur.

40
 AIML for Mechanical Engineers

Advantages:
Interpretability: Linear regression models are easy to interpret, making it straightforward to
understand the relationship between the independent and dependent variables.
Computational Efficiency: Linear regression models are computationally efficient and can
handle large datasets with ease.
Versatility: Linear regression can be applied to both continuous and categorical variables,
making it suitable for a wide range of applications.
Baseline Model: Linear regression serves as a baseline model for more complex machine
learning algorithms, providing a simple and interpretable benchmark for comparison.

Challenges:
Assumption Violations: Linear regression relies on several assumptions that may not hold true
in real-world datasets, such as the linearity and independence of variables.
Overfitting/Underfitting: Linear regression may suffer from underfitting (when the model is
too simple to capture the underlying patterns) or overfitting (when the model is too complex
and captures noise in the data).
Limited Complexity: Linear regression assumes a linear relationship between the variables,
which may not always be the case in practice.

In summary, linear regression is a powerful and widely used technique for modelling the
relationship between variables. By understanding its principles, assumptions, and applications,
you can leverage linear regression to gain insights from data and make informed decisions in
various domains.

Department of Mechanical Engineering, DYPCET, Kolhapur.

41
 AIML for Mechanical Engineers

Stages in Linear Regression Analysis

Linear regression analysis is a systematic approach used to understand and model the
relationship between one or more independent variables and a dependent variable. Linear
Regression Analysis consists of more than just fitting a linear line through a cloud of data
points. It consists of 3 stages

1. Analysing the Correlation and Directionality of the Data:

Before fitting a linear model, it's essential to analyse the correlation and directionality
of the data. This involves examining the relationships between the independent and dependent
variables. Correlation analysis helps determine the strength and direction of the linear
relationship between variables. It quantifies the degree to which changes in one variable are
associated with changes in another variable.
Directionality analysis examines whether the relationship between variables is positive
or negative. A positive correlation indicates that as one variable increases, the other variable
also tends to increase, while a negative correlation indicates the opposite.
Scatter plots, correlation coefficients (such as Pearson's correlation coefficient), and
hypothesis testing (such as t-tests) are commonly used techniques for analyzing correlation and
directionality.
2. Estimating the Model - Fitting the Line:
Once the correlation and directionality of the data are understood, the next stage
involves estimating the linear model by fitting a line to the data.
In simple linear regression, a straight line is fitted to the data using the method of least squares.
The goal is to minimize the sum of squared differences between the observed values and the
values predicted by the model.
3. Evaluating the Validity and Usefulness of the Model:
Once the model is estimated, it's important to evaluate its validity and usefulness to
ensure that it provides meaningful insights and predictions.
Model evaluation involves assessing several aspects, including:
Goodness of fit: How well does the model fit the data? This can be evaluated using metrics
such as R-squared (coefficient of determination) and adjusted R-squared.
Residual analysis: Are the residuals (the differences between observed and predicted values)
normally distributed and homoscedastic? Residual plots and statistical tests (such as Shapiro-
Wilk test) can be used to assess this.
Significance of coefficients: Are the coefficients statistically significant? Hypothesis testing
(such as t-tests) can be used to determine whether the coefficients are different from zero.
Finally, the model should be validated using external data or compared to alternative models
to ensure its robustness and generalizability.

Department of Mechanical Engineering, DYPCET, Kolhapur.

42
 AIML for Mechanical Engineers

1. Use a Simple Model That Fits Many Models

This principle suggests that it's often preferable to start with a simpler linear regression
model and gradually increase complexity only if necessary. Simpler models are generally easier
to interpret and less prone to overfitting the data.
2. Consider Variable Transformations
Linear regression assumes a linear relationship between the independent and dependent
variables. The image suggests that you may need to transform variables if the relationship is
not linear. Common transformations include taking the logarithm, square root, or using
exponential functions.
3. Speed Up Your Computations and Make Them More Reliable
The advice here is to leverage statistical software packages like R or Python libraries like
Scikit-learn to perform linear regression analysis. These tools can automate calculations and
ensure accuracy.
4. Learn Regression Methods Through Live Examples
The image emphasizes the importance of practical experience and working through real-
world examples to gain a deeper understanding of linear regression concepts.
5. Consider Five Key Assumptions Concerning Data
regression relies on several assumptions about the data for accurate results. These
assumptions are:
▪ Linearity: There is a linear relationship between the independent and dependent
variables.
▪ Independence: The errors in the data are independent of each other.
▪ Homoscedasticity: The variance of the errors is constant across all levels of the
independent variable.
▪ Normality: The errors are normally distributed.
▪ No Multicollinearity: The independent variables are not highly correlated with each
other.
It's important to assess these assumptions before performing linear regression analysis and
address any violations to ensure the validity of your results.

Department of Mechanical Engineering, DYPCET, Kolhapur.

43
 AIML for Mechanical Engineers

Application of Linear Regression

Linear regression, with its simplicity and interpretability, finds diverse applications across
various domains. Here are some notable applications:
1. Market Analysis and Sales Maximization:
• Linear regression is extensively used in market analysis to understand consumer
behaviour and optimize marketing strategies.
• By analyzing historical sales data along with marketing efforts such as advertising
expenditure, promotions, and pricing strategies, businesses can identify trends and
patterns that influence sales.
• Linear regression models can predict the impact of different marketing strategies on
sales and help businesses allocate resources effectively to maximize revenue.
• For example, a retail company can use linear regression to analyze the effectiveness of
different advertising campaigns and promotional offers in boosting sales.
2. Financial Study and Operational Performance Evaluation:
• Linear regression is employed in financial studies to evaluate the operational
performance of establishments, such as companies, banks, and investment portfolios.
• By analyzing financial indicators such as revenue, expenses, profits, and stock prices,
linear regression models can assess the financial health and stability of an entity.
• Financial analysts use linear regression to identify factors that influence financial
performance, such as market conditions, competition, and macroeconomic trends.
• For instance, linear regression can be used to analyze the relationship between a
company's marketing expenditure and its profitability, helping managers make
informed decisions about resource allocation and investment strategies.
3. Sports Analysis and Game Attendance Prediction:
• Linear regression is applied in sports analytics to predict various outcomes, including
game attendance, team performance, and player salaries.
• By analyzing factors such as team status, market size, historical game attendance, and
ticket prices, linear regression models can forecast the expected attendance at sporting
events.
• Sports organizations use these predictions to optimize ticket pricing, stadium capacity,
and marketing efforts to maximize revenue and fan engagement.
• For example, a sports franchise can use linear regression to analyze the impact of team
performance, weather conditions, and competing events on game attendance, allowing
them to make strategic decisions about scheduling and promotions.

Department of Mechanical Engineering, DYPCET, Kolhapur.

44
 AIML for Mechanical Engineers

The image above is a conceptual diagram outlining the linear regression process.

Data Collection: This initial step involves gathering the data that will be used to train the linear
regression model. The data should be relevant to the problem you're trying to solve and include
independent variables (predictor variables) and a dependent variable (target variable).

• Data Cleaning and Preprocessing:

• Once you have the data, you'll likely need to clean and prepare it before using it in your
model. This may involve tasks like:
• Identifying and handling missing values
• Dealing with outliers
• Encoding categorical variables
• Normalizing or scaling numerical features
• Exploratory Data Analysis (EDA):
• Exploratory data analysis involves getting a better understanding of the data you're
working with. This might involve:
• Visualizing the distribution of your variables using histograms, scatter plots, and
boxplots
• Calculating summary statistics like mean, median, and standard deviation
• Identifying potential relationships between variables
• Model Fitting
• In this stage, you'll define and fit the linear regression model to your data. This
involves specifying the relationship between the independent and dependent variables
using a linear equation. The model will learn the coefficients of this equation by
minimizing the difference between the predicted values and the actual values in your
data.
• Model Evaluation: After fitting the model, you need to assess its performance. Common
metrics used for evaluating linear regression models include:
o R-squared: This metric represents the proportion of variance in the dependent
variable that can be explained by the independent variables.

Department of Mechanical Engineering, DYPCET, Kolhapur.

45
 AIML for Mechanical Engineers

o Mean squared error (MSE): This metric measures the average squared
difference between the predicted values and the actual values.
o Root mean squared error (RMSE): The square root of the MSE.
• Model Refinement: Based on the evaluation results, you may need to refine your model.
This could involve:
o Trying different transformations of the variables
o Selecting a different model complexity (e.g., adding or removing terms from
the linear equation)
o Collecting more data
• Prediction: Once you have a well-performing model, you can use it to make predictions
on new data. For instance, you can predict the value of the dependent variable for new
observations based on the values of the independent variables.

The arrows in the diagram suggest that linear regression is an iterative process. You may need
to go back and forth between different steps as you clean your data, explore relationships,
evaluate your model, and refine it.

Multi Linear Regression

Multilinear regression, also known as multiple linear regression, is an extension of simple

linear regression that allows for the modelling of relationships between a dependent variable
and multiple independent variables. It is a powerful statistical technique used to analyze and
predict outcomes when multiple predictors influence the response variable.

Key Concepts:

Multiple Independent Variables:

In multilinear regression, there are two or more independent variables (also called
predictors or features) that are used to predict a single dependent variable.
Each independent variable has its own coefficient, representing the strength and direction of its
relationship with the dependent variable.

Linear Relationship:
Multilinear regression assumes a linear relationship between the independent variables and the
dependent variable. The relationship is represented by a linear equation of the form:

Department of Mechanical Engineering, DYPCET, Kolhapur.

46
 AIML for Mechanical Engineers

Applications of Multi Linear Regression:

Economics: Analysing the factors affecting GDP growth, inflation rates, and unemployment
rates.
Marketing: Predicting sales based on advertising expenditure, pricing strategies, and market
demographics.
Finance: Modelling stock prices based on economic indicators, company fundamentals, and
market sentiment.
Healthcare: Predicting patient outcomes based on medical history, treatment variables, and
demographic factors.
Environmental Science: Analysing the impact of environmental factors on climate change, air
quality, and biodiversity.

Advantages:
Multiple Predictors: Multilinear regression allows for the analysis of the combined effects of
multiple predictors on the dependent variable.
Interpretability: The coefficients in the multilinear regression model provide insights into the
strength and direction of the relationships between the independent variables and the dependent
variable.
Flexibility: Multilinear regression can accommodate both quantitative and categorical
independent variables, making it suitable for a wide range of data types.

Challenges:
Assumption Violations: Multilinear regression relies on assumptions such as linearity,
independence, homoscedasticity, and normality of residuals. Violations of these assumptions
can affect the validity of the model.
Overfitting/Underfitting: Multilinear regression models may suffer from overfitting (when
the model is too complex and captures noise in the data) or underfitting (when the model is too
simple to capture the underlying patterns).

Department of Mechanical Engineering, DYPCET, Kolhapur.

47
 AIML for Mechanical Engineers

Logistic Regression
Logistic regression is a process of modelling the probability of a discrete outcome given
an input variable. The most common logistic regression models a binary outcome; something
that can take two values such as true/false, yes/no, and so on.
Logistic regression is a powerful statistical technique used for modelling the
relationship between a binary dependent variable and one or more independent variables.
Despite its name, logistic regression is a classification algorithm rather than a regression
algorithm, as it is commonly used to predict the probability of occurrence of a binary outcome.

Time Click Add Time Click Add

20 No 50 YES
40 No 45 YES
10 YES 15 No
30 No 35 YES
60 No 85 No

Key Concepts:
Binary Dependent Variable:
Logistic regression is used when the dependent variable (or response variable) is categorical
with two possible outcomes, typically coded as 0 and 1.
Examples of binary outcomes include yes/no, success/failure, pass/fail, and presence/absence.

Odds Ratio:
Logistic regression models the odds of the positive outcome rather than the outcome
itself. The odds ratio represents the odds of the positive outcome occurring relative to the odds
of the negative outcome. It is calculated as the ratio of the probability of the positive outcome
to the probability of the negative outcome.
Maximum Likelihood Estimation (MLE):
The parameters (coefficients) of the logistic regression model are estimated using
maximum likelihood estimation (MLE), a statistical method that maximizes the likelihood of
observing the given data under the assumed model. MLE finds the values of the coefficients
that make the observed data most probable given the assumed logistic regression model.

Department of Mechanical Engineering, DYPCET, Kolhapur.

48
 AIML for Mechanical Engineers

In summary, logistic regression is a versatile and widely used classification technique for
modelling the relationship between independent variables and a binary outcome. By
understanding its principles, applications, advantages, and challenges, analysts can leverage
logistic regression to make accurate predictions and informed decisions in various domains.

The image above is a diagram of a logistic regression model. Here's a detailed breakdown of
the components in the image:

S-shaped curve: This curve represents the logistic function, which is the core of logistic
regression. It maps the linear combination of the independent variables (x-axis) to a probability
value between 0 and 1 (y-axis). The S-shape ensures that the predicted probabilities always fall
within this range.

Independent variables (X): These are the predictor variables that you input into the model.
The model will learn the relationship between these variables and the dependent variable.

Linear combination of weights (ΣwX): This represents the weighted sum of the independent
variables, where each weight (w) reflects the model's learned importance of each variable in
predicting the outcome.

Threshold (θ): The threshold is a decision boundary on the x-axis of the S-shaped curve.
Values above the threshold are classified as positive outcomes (1), while values below the
threshold are classified as negative outcomes (0).

Predicted probability (y): This is the output of the logistic regression model, representing the
likelihood of a positive outcome (between 0 and 1) for a given set of independent variables.

Here's how logistic regression works:

1. Data Collection: You start by collecting data that includes the independent variables and
the dependent variable (binary outcome).
2. Model Training: The logistic regression model is trained on the data. During training, the
model learns the weights for each independent variable and the threshold value. This is
done by minimizing the difference between the predicted probabilities and the actual
outcomes in the data.

Department of Mechanical Engineering, DYPCET, Kolhapur.

49
 AIML for Mechanical Engineers

3. Prediction: Once trained, you can use the model to predict the probability of a positive
outcome for new data points. The model takes the new data point's independent variable
values as input, calculates the weighted sum, and then applies the sigmoid function to get
the predicted probability.

Types of Logistic Regression

There are three main types of logistic regression: binary, multinomial and ordinal. They differ
in execution and theory. Binary regression deals with two possible values, essentially: yes or
no. Multinomial logistic regression deals with three or more values.

1. Binary Logistic Regression:

• Binary logistic regression is the most common type and is used when the dependent
variable has only two possible outcomes, typically coded as 0 and 1.
• It models the probability of occurrence of one of the two outcomes based on one or more
independent variables.
• Example applications include predicting whether a customer will churn (1) or not churn
(0), or whether a patient has a disease (1) or does not have the disease (0).
2. Multinomial Logistic Regression:
• Multinomial logistic regression is used when the dependent variable has three or more
categorical outcomes that are not ordered.
• It models the probabilities of each category of the dependent variable relative to a reference
category, often referred to as the baseline category.
• Example applications include predicting the likelihood of a person choosing one of several
options, such as predicting which type of product a customer will purchase from a set of
options.
3. Ordinal Logistic Regression:
• Ordinal logistic regression is used when the dependent variable has three or more ordered
categorical outcomes.
• It models the cumulative probabilities of the ordered categories of the dependent variable.
• Example applications include predicting the likelihood of a person choosing one of several
ordered options, such as predicting the satisfaction level (low, medium, high) of customers
based on service quality.

Department of Mechanical Engineering, DYPCET, Kolhapur.

50
 AIML for Mechanical Engineers

Recommender System using Logistic Regression

The image above is a conceptual diagram outlining the workflow of a recommender system.
Recommender systems are algorithms that suggest items (products, movies, music, etc.) to
users based on their preferences and past behaviour. Here's a breakdown of the key stages
depicted in the image:

1. User Data Collection

This initial stage involves gathering data about users. This data can include explicit
feedback (e.g., ratings, reviews, thumbs up/down) or implicit feedback (e.g., purchase
history, browsing behaviour, clicks).
2. Item Data Collection
Information about the items themselves is also collected. This may include product
descriptions, attributes (e.g., genre, category, brand), or content features.
3. Data Preprocessing
Both user and item data may need cleaning and preparation before use. This could involve
handling missing values, scaling numerical features, or encoding categorical variables.
4. Model Selection & Training
Several machine learning algorithms can be used for recommender systems, including
collaborative filtering, content-based filtering, or hybrid approaches that combine both
techniques. The choice of algorithm depends on the nature of the data and the specific
recommendation task.
5. Model Evaluation
Once trained, the recommender system's performance is evaluated on a held-out test
dataset. This helps assess how well the model can predict user preferences for unseen items.
6. Recommendations
When a user interacts with the system, the model generates personalized recommendations
based on the user's profile and past behaviour. The recommendations are typically ranked
by predicted preference, with the most relevant items appearing at the top.
7. Feedback Loop
User interaction with the recommendations provides implicit feedback that can be fed back
into the system to improve its accuracy over time. This feedback loop is crucial for
recommender systems to continuously learn and adapt to evolving user preferences.

Department of Mechanical Engineering, DYPCET, Kolhapur.

51
 AIML for Mechanical Engineers

Advantages of Logistic Regression:

1. Easier Implementation of Machine Learning Methods:
▪ Logistic regression is relatively simple to implement and interpret compared to more
complex machine learning algorithms.
▪ It requires minimal data preprocessing and parameter tuning, making it an excellent
choice for beginners and for scenarios where transparency and interpretability are
important.
▪ With the availability of libraries and packages in various programming languages,
setting up a logistic regression model for classification tasks is straightforward and
requires minimal coding effort.
2. Suitable for Linearly Separable Datasets
▪ Logistic regression performs well when the data is linearly separable, meaning a
straight line can effectively separate the two classes in the feature space.
▪ In cases where the relationship between the independent variables and the log-odds of
the dependent variable is approximately linear, logistic regression can provide accurate
predictions
▪ Linear separability simplifies the decision boundary, making it easier to interpret and
understand the model's predictions.
3. Robustness to Noise and Irrelevant Features:
▪ Logistic regression is less susceptible to overfitting compared to more complex
models, especially when the number of features is small relative to the number of
observations.
▪ It can handle noisy data and irrelevant features without significantly impacting its
performance, making it suitable for datasets with high-dimensional feature spaces
4. Probability Estimation:
▪ Logistic regression models provide probabilistic outputs, representing the probability
of belonging to a particular class.
▪ This probabilistic interpretation allows for more nuanced decision-making and
uncertainty quantification compared to deterministic classifiers.
▪ Decision thresholds can be adjusted based on the specific needs of the application,
balancing trade-offs between false positives and false negatives
5. Interpretability:
• Logistic regression models offer straightforward interpretability, as the coefficients
associated with each independent variable indicate the strength and direction of their
influence on the predicted outcome.
• Interpretability is essential for understanding the factors driving the predictions and
gaining insights into the relationships between variables.

Department of Mechanical Engineering, DYPCET, Kolhapur.

52
 AIML for Mechanical Engineers

Applications of Logistic Regression

• Biomedical Research: Predicting the likelihood of disease occurrence based on risk
factors such as age, gender, and genetic markers.
• Marketing: Predicting customer churn, determining the likelihood of response to
marketing campaigns, and segmenting customers based on purchasing behaviour.
• Finance: Assessing credit risk, predicting loan default, and detecting fraudulent
transactions.
• Social Sciences: Analysing survey data, predicting voting behaviour, and studying the
impact of interventions on behaviour change.

Linear Regression vs Logistic Regression

Department of Mechanical Engineering, DYPCET, Kolhapur.

53
 AIML for Mechanical Engineers

The image above is a diagram showing a high-level overview of the steps involved in building
a machine learning model.
1. Problem Definition
This initial stage involves clearly defining the problem you're trying to solve with a
machine learning model. Formulating a well-defined problem statement is crucial for the
success of the machine learning project, as it guides the entire process from data collection to
model selection and evaluation.
2. Data Collection
Once you have a clear understanding of the problem, you need to gather the data that
will be used to train the machine learning model. The data should be relevant to the problem
and include features (variables) that can potentially influence the outcome you want to predict.
Data can be collected from various sources such as databases, surveys, web scraping, or APIs.
3. Data Understanding
After collecting the data, it's essential to explore and analyze it to gain insights into its
properties, distribution, and potential challenges. This stage involves data cleaning tasks like
identifying and handling missing values, outliers, and inconsistencies. Exploratory data
analysis (EDA) may also involve visualizing the data using histograms, scatter plots, and box
plots to understand the relationships between variables.
4. Data Preparation
In this stage, the data is preprocessed to prepare it for use in a machine learning model.
Preprocessing steps may involve:
o Scaling: Scaling numerical features to a common range to prevent features with larger
values from dominating the model during training.
o Encoding: Transforming categorical variables into numerical representations that machine
learning models can understand. Common encoding techniques include one-hot encoding
and label encoding.
o Feature Engineering: Creating new features from existing ones that might be more
informative for the model.

Department of Mechanical Engineering, DYPCET, Kolhapur.

54
 AIML for Mechanical Engineers

5. Model Selection
Here, you choose a suitable machine learning algorithm for your task. The choice of
algorithm depends on several factors, including the nature of the problem (classification,
regression, etc.), the type of data you have (tabular, image, text), and the desired outcome.
Common machine learning algorithms include linear regression, decision trees, random forests,
support vector machines, and deep learning models like convolutional neural networks (CNNs)
for image recognition or recurrent neural networks (RNNs) for sequence data.
6. Model Training
In this stage, the machine learning model is trained on the prepared data. The training
process involves fitting the model to the data and learning the patterns and relationships
between the features and the target variable. Training algorithms iteratively adjust the model's
parameters to minimize the error between the model's predictions and the actual values in the
training data.
7. Model Evaluation
After training the model, it's crucial to evaluate its performance on a separate test
dataset to assess its generalizability and avoid overfitting. Overfitting refers to a scenario where
the model performs well on the training data but poorly on unseen data. Common metrics used
for evaluation depend on the machine learning task, but they may include accuracy, precision,
recall, F1-score, mean squared error (MSE), or root mean squared error (RMSE).
8. Model Deployment
If the model performs well on the test dataset, you can deploy it for real-world use.
Deployment involves integrating the model into a production environment where it can make
predictions on new data. For instance, the model could be deployed as a web service or
embedded into a mobile application.
9. Model Monitoring
Once deployed, it's essential to monitor the model's performance over time to ensure it
continues to make accurate predictions. Monitoring may involve tracking the model's
evaluation metrics on new data and retraining the model with fresh data if its performance
degrades.

Department of Mechanical Engineering, DYPCET, Kolhapur.

55
 AIML for Mechanical Engineers

Unit IV
Introduction to
Power BI

Department of Mechanical Engineering, DYPCET, Kolhapur.

56
 AIML for Mechanical Engineers

Power BI: Transforming Data into Insights

Introduction:
Power BI is a powerful business analytics tool developed by Microsoft that enables users to
visualize and analyse data from various sources, gain insights, and make data-driven decisions.
It offers a suite of features for data preparation, modelling, visualization, and collaboration,
making it a popular choice for businesses of all sizes.

Development:
Power BI was first introduced by Microsoft in 2013 as a cloud-based business intelligence
service.
Over the years, it has evolved into a comprehensive suite of tools, including Power BI Desktop
(for authoring reports), Power BI Service (for sharing and collaborating on reports), and Power
BI Mobile (for accessing reports on mobile devices).
Microsoft continues to invest in Power BI, regularly releasing updates and new features to
enhance functionality and usability.

Key Features of Power BI:

Data Connectivity:
Power BI supports connectivity to a wide range of data sources, including Excel spreadsheets,
SQL databases, cloud services (such as Azure, Google Analytics, and Salesforce), and on-
premises data sources. Users can import data directly into Power BI or establish live
connections to maintain real-time access to data.
Data Preparation:
Power BI offers robust data preparation capabilities, allowing users to clean, transform, and
shape data using a user-friendly interface. Features such as data profiling, data cleansing, and
data modelling simplify the process of preparing data for analysis.
Data Modelling:
Users can create relationships between different datasets, define calculated columns and
measures, and build data models using Power BI's intuitive modelling tools. Power BI's Data
Analysis Expressions (DAX) language enables advanced calculations and aggregation
functions for creating sophisticated analytical models.
Visualization:
Power BI provides a rich set of visualization options, including bar charts, line graphs, pie
charts, maps, and custom visuals. Users can customize and format visualizations to suit their
preferences, adding titles, labels, colors, and interactive elements to enhance readability and
engagement.
Dashboarding:
Power BI allows users to create interactive dashboards by pinning visualizations from multiple
reports onto a single canvas. Dashboards can be customized with filters, slicers, and other
interactive controls to enable users to explore data dynamically.
Collaboration and Sharing:
Power BI Service enables users to publish, share, and collaborate on reports and dashboards
with colleagues within or outside their organization. Features such as content packs, apps, and
workspaces facilitate seamless collaboration and knowledge sharing across teams.

Department of Mechanical Engineering, DYPCET, Kolhapur.

57
 AIML for Mechanical Engineers

Data Summarization and Data Visualization:

Data Summarisation:
Data summarization involves condensing large volumes of data into a concise and
understandable format. Techniques such as aggregation, grouping, and statistical measures like
mean, median, mode, and standard deviation are used to summarize data.
Summarized data provides a high-level overview of trends, patterns, and outliers, facilitating
easier interpretation and analysis.

Data Visualization:
Data visualization is the graphical representation of data using charts, graphs, maps, and other
visual elements. Visualization enables users to explore and understand complex datasets more
effectively by presenting information in a visually intuitive manner.
Common types of data visualizations include bar charts, line graphs, pie charts, scatter plots,
heatmaps, and geospatial maps. Interactive visualization tools allow users to manipulate and
explore data dynamically, enabling deeper insights and analysis.

Benefits:
Enhances Understanding: Visual representations of data make it easier to understand
complex relationships and patterns, even for non-technical users.
Facilitates Decision-Making: Visualizations provide actionable insights that support decision-
making processes, helping stakeholders identify trends, opportunities, and areas for
improvement.
Improves Communication: Visualizations serve as a powerful communication tool, enabling
stakeholders to share insights and findings with others more effectively.
Supports Data Exploration: Interactive visualizations allow users to explore data
dynamically, drilling down into specific details and uncovering hidden insights.

Applications:
➢ Power BI finds applications across various industries and functions, including:
➢ Business Intelligence and Analytics
➢ Financial Reporting and Analysis
➢ Sales and Marketing Analytics
➢ Operations and Supply Chain Management
➢ Human Resources and Talent Management

Advantages:

User-Friendly Interface: Power BI offers an intuitive interface with drag-and-drop

functionality, making it accessible to users with varying levels of technical expertise.
Integration: Power BI integrates seamlessly with other Microsoft products and services, such
as Excel, SharePoint, and Azure, enhancing productivity and collaboration.
Scalability: Power BI is highly scalable, capable of handling large volumes of data and
supporting complex analytical models and visualizations.
Cost-Effective: Power BI offers flexible pricing options, including free and subscription-based
plans, making it an affordable solution for businesses of all sizes.

Department of Mechanical Engineering, DYPCET, Kolhapur.

58
 AIML for Mechanical Engineers

Building blocks of Power BI

The basic building blocks in Power BI are:
• Visualizations
• Datasets
• Reports
• Dashboards
• Tiles

1. Visualizations
A visualization is a representation of data in a visual format. It could be a line chart, a bar
graph, a color coded map or any visual way to present the data.

Visualizations can be a simple number representing a significant calculation or it could be

more complex like multiple charts showing the proportion of users participating in a survey.
The main idea of visualisation is to show the data in a way that tells the story that is lying
underneath it. Like the saying goes: a picture says a thousand words.

2. Datasets
A dataset is a collection of data that Power BI uses to create its visualizations. You can
have a simple dataset that’s based on a single table from a Microsoft Excel workbook,
similar to what’s shown in the following image.
Dataset can also be a combination of many different sources, which can be filtered
using Power BI and combined into one to use.

Department of Mechanical Engineering, DYPCET, Kolhapur.

59
 AIML for Mechanical Engineers

For example: One data source contains countries and locations in the form of latitude and
longitude. Another data source contains demographics of these countries like population and
GDP. Power BI can combine these two data sources into one dataset which can be used for
visualizations.
An important feature of Power BI is the ability to connect to various data sources
using its connectors. Whether the data you want is in Excel or a Microsoft SQL Server
database, in Azure or Oracle, or in a service like Facebook, Salesforce, or MailChimp, Power
BI has built-in data connectors that let you easily connect to that data, filter it if necessary,
and bring it into your dataset.
After you have a dataset, you can begin creating visualizations that show different
portions of it in different ways, and gain insights based on what you see. That is where
reports come in.

3. Reports
In Power BI, a Report is a collection of visualizations that appear together on one
or more pages. A report in Power BI is a collection of items that are related to each other.
We will be working with the gapminder data to create the report below that looks at the
GDP, population and life expectancy by global regions.

Reports let us create and structure visualizations on pages based on the way the we want to
tell the story.

4. Dashboards
A Power BI dashboard is a collection of visuals from a single page that you can share with
others. Often it is a selected group of visuals that provide quick insight into the data or story
you are trying to present.
A dashboard must fit on a single page, often called a canvas (the canvas is the blank
backdrop in Power BI Desktop or the service, where you put visualizations). Think of it like
the canvas that an artist or painter uses — a workspace where you create, combine, and
rework interesting and compelling visuals. You can share dashboards with other users or
groups, who can then interact with your dashboards when they’re in the Power BI service
or on their mobile device.

Department of Mechanical Engineering, DYPCET, Kolhapur.

60
 AIML for Mechanical Engineers

Dashboards in Power BI: Visualizing Insights for Actionable Intelligence

Introduction:
Dashboards in Power BI serve as dynamic, interactive canvases for presenting key insights and
performance metrics derived from data analysis. They enable users to monitor, analyze, and
share real-time information in a visually compelling and intuitive manner.

Development:
Power BI Dashboards were introduced as part of the Power BI suite by Microsoft to
enhance data visualization and decision-making capabilities. Continual updates and
enhancements to Power BI have refined dashboard creation, offering users an ever-expanding
array of features and capabilities.

Key Features of Power BI Dashboards:

Visualizations:
Power BI Dashboards support a wide range of visualizations, including charts, graphs,
maps, and custom visuals, allowing users to represent data in a meaningful and impactful way.
Users can select from a library of pre-built visualizations or create custom visualizations using
Power BI's developer tools.

Interactivity:
Dashboards in Power BI are highly interactive, enabling users to explore data
dynamically by applying filters, drill-downs, and slicers to visualize specific insights.
Interactive elements such as buttons, bookmarks, and tooltips further enhance user engagement
and exploration.
Real-Time Updates:
Power BI Dashboards can display real-time data updates, ensuring that users have
access to the most current information available. Live connections to data sources enable
automatic refreshes and ensure that dashboards reflect the latest changes in underlying data.
Mobile Accessibility:
Power BI Dashboards are optimized for mobile devices, allowing users to access and
interact with dashboards on smartphones and tablets using the Power BI Mobile app.

Department of Mechanical Engineering, DYPCET, Kolhapur.

61
 AIML for Mechanical Engineers

Responsive design ensures that dashboards adapt to different screen sizes and orientations for
a seamless mobile experience.
Customization:
Users can customize dashboards in Power BI to suit their specific needs and
preferences, including layout, colors, branding, and interactive features.
Custom visuals and themes enable users to create unique and visually stunning dashboards
tailored to their audience.

Applications of Power BI Dashboards:

Dashboards in Power BI find applications across various industries and business functions,
including:
Sales and Marketing: Tracking sales performance, pipeline management, and marketing
campaign effectiveness.
Finance and Accounting: Monitoring financial metrics, budget vs. actuals, and cash flow
analysis.
Operations and Supply Chain: Managing inventory levels, production schedules, and
supplier performance.
Human Resources: Analyzing workforce demographics, employee engagement, and talent
acquisition metrics.

Advantages:
Actionable Insights: Power BI Dashboards provide users with actionable insights derived
from data analysis, enabling informed decision-making and driving business outcomes.
Visualization: Visualizations in Power BI Dashboards make complex data more accessible and
understandable, facilitating communication and collaboration across teams.
Interactivity: Interactive features empower users to explore data dynamically, uncovering
hidden patterns and trends that drive deeper understanding and analysis.
Real-Time Monitoring: Real-time updates ensure that dashboards reflect the latest
information, enabling timely decision-making and response to changing business conditions.

Steps to create dashboards in Power BI:

Step 1: Prepare Your Data

Start by gathering and preparing the data you want to visualize in your dashboard. This
may involve importing data from various sources, such as Excel files, databases, or cloud
services.
Clean and transform the data as needed to ensure accuracy and consistency. Use Power Query
Editor to perform data cleaning, shaping, and modelling operations.
Step 2: Create Visualizations
Open Power BI Desktop and load your prepared data into the report canvas.
Use the Fields pane to drag and drop fields onto the canvas to create visualizations such as
charts, graphs, tables, and maps. Customize each visualization by adjusting properties such as
colors, labels, axes, and data labels to enhance readability and visual appeal.
Step 3: Design the Dashboard Layout
After creating individual visualizations, arrange them on the report canvas to create a
cohesive dashboard layout. Use the layout tools in Power BI Desktop to organize visualizations
into rows, columns, or groups, and adjust the size and position of each visualization as needed.

Department of Mechanical Engineering, DYPCET, Kolhapur.

62
 AIML for Mechanical Engineers

Consider the flow of information and prioritize the most important insights to be displayed
prominently on the dashboard.
Step 4: Add Interactivity
Enhance your dashboard with interactive features such as slicers, filters, and drill-
downs to enable users to explore the data dynamically.
Use slicers to filter data based on specific criteria, such as time periods, regions, or product
categories. Enable cross-filtering and cross-highlighting between visualizations to maintain
context and interactivity.
Step 5: Create Dashboard Tiles
Once you have designed your dashboard layout and added interactivity, you can convert
individual visualizations into dashboard tiles.
Select the visualizations you want to include on the dashboard and click the "Pin Live Page"
or "Pin to Dashboard" button to pin them to the dashboard.
Choose the dashboard to which you want to pin the visualization tiles or create a new
dashboard.
Step 6: Customize Dashboard Tiles
After pinning visualizations to the dashboard, customize the dashboard tiles to optimize
their appearance and functionality.
Resize and rearrange tiles to create a visually appealing layout that fits the dashboard canvas.
Add titles, subtitles, and descriptions to provide context and guidance for users navigating the
dashboard.
Step 7: Publish and Share the Dashboard
Once you have finalized your dashboard design, publish it to the Power BI Service to
share it with others. Click the "Publish" button in Power BI Desktop to upload the dashboard
to the Power BI Service. Share the dashboard with colleagues or stakeholders by granting them
access permissions or embedding it in SharePoint, Teams, or other applications.
Step 8: Monitor and Maintain the Dashboard
Regularly monitor the performance and usage of your dashboard in the Power BI
Service. Use usage metrics and feedback from users to identify areas for improvement and
optimization. Update the dashboard as needed to reflect changes in data or business
requirements, ensuring that it remains relevant and valuable to users.

Power BI Dashboards vs Reports

Capability Dashboards Reports

Pages Has Only One Page Has one or more pages
Data Sources Contains one or more Contains a single dataset per
reports and one or more report
datasets per dashboard
Filtering It can’t filter or slice the It can filter, highlight and
view slice the view in various
ways
Availability on Power BI Dashboards are not available Can create reports on Power
Desktop in Power BI Desktop BI Desktop
Featured You can set one dashboard Cannot create a featured
as your "featured" dashboard report

Department of Mechanical Engineering, DYPCET, Kolhapur.

63
 AIML for Mechanical Engineers

Unit V
Advanced Excel
Techniques

Department of Mechanical Engineering, DYPCET, Kolhapur.

64
 AIML for Mechanical Engineers

Microsoft Excel: An Overview

Microsoft Excel is a powerful spreadsheet software developed by Microsoft Corporation,

widely used for data analysis, calculation, visualization, and reporting. It provides a user-
friendly interface with a range of features designed to facilitate various tasks, from basic
arithmetic calculations to complex data modelling and visualization. Here's an overview of
Microsoft Excel:

Introduction:

Microsoft Excel is part of the Microsoft Office suite of productivity software, which also
includes Word, PowerPoint, Outlook, and Access.
It was first released in 1985 for the Apple Macintosh and later for Microsoft Windows in 1987,
becoming one of the most popular spreadsheet applications globally. Excel uses a grid of cells
organized into rows and columns, where users can enter data, perform calculations, and create
charts and graphs.

Key Features of MS Excel:

Data Entry and Formatting:

Users can enter and manipulate data in individual cells, applying various formatting options
such as font styles, colors, and borders to enhance readability. Excel supports a wide range of
data types, including numbers, text, dates, and formulas, allowing for versatile data
representation and analysis.
Formulas and Functions:
Excel offers a vast library of built-in functions and formulas for performing mathematical,
statistical, financial, and logical operations. Users can create custom formulas using operators
and functions to automate calculations and analyse data more efficiently.
Charts and Graphs:
Excel provides powerful charting tools to create various types of charts and graphs, including
bar charts, line graphs, pie charts, and scatter plots.
Users can customize chart elements such as axes, titles, legends, and data labels to visualize
data trends and patterns effectively.
Data Analysis Tools:

Department of Mechanical Engineering, DYPCET, Kolhapur.

65
 AIML for Mechanical Engineers

Excel includes advanced data analysis tools such as PivotTables, PivotCharts, and Data
Analysis ToolPak, which enable users to summarize, filter, and analyze large datasets quickly.
These tools facilitate tasks such as trend analysis, regression analysis, and hypothesis testing,
helping users gain insights from their data.
Data Visualization:
Excel allows users to create visually appealing dashboards and reports by combining charts,
graphs, and tables on a single worksheet.
Users can use features like conditional formatting and sparklines to highlight key trends and
outliers in their data visually.

Applications:

Microsoft Excel is widely used across various industries and sectors for diverse purposes,
including:
➢ Financial Analysis and Budgeting
➢ Business Planning and Forecasting
➢ Data Management and Reporting
➢ Project Management and Task Tracking
➢ Academic Research and Statistical Analysis
➢ Business Intelligence: Visualizations are widely used in business intelligence for
monitoring key performance indicators (KPIs), analyzing trends, and identifying areas for
optimization.
➢ Data Analytics: Data visualizations aid in exploratory data analysis, hypothesis testing,
and model evaluation, enabling data scientists to derive insights and validate findings.
➢ Presentations and Reports: Visualizations enhance the presentation of findings in reports,
presentations, and dashboards, making information more engaging and accessible to
audiences.

Advantages:
User-Friendly Interface: Excel's intuitive interface makes it accessible to users with varying
levels of expertise.
Versatility: Excel can handle a wide range of tasks, from simple calculations to complex data
analysis and modelling.
Integration: Excel integrates seamlessly with other Microsoft Office applications and third-
party software, enhancing productivity and collaboration.
Cost-Effective: Excel is widely available and relatively inexpensive compared to specialized
data analysis software.

Department of Mechanical Engineering, DYPCET, Kolhapur.

66
 AIML for Mechanical Engineers

Features of MS Excel

Ribbon
Th Ribbon in MS-Excel is the topmost row of tabs that provide the user with different
facilities/functionalities. These tabs are:

Home Tab
It provides the basic facilities like changing the font, size of text, editing the cells in the
spreadsheet, autosum, etc.

Insert Tab
It provides the facilities like inserting tables, pivot tables, images, clip art, charts, links, etc.

Page layout
It provides all the facilities related to the spreadsheet-like margins, orientation, height, width,
background etc. The worksheet appearance will be the same in the hard copy as well.

Formulas
It is a package of different in-built formulas/functions which can be used by user just by
selecting the cell or range of cells for values.

Data
The Data Tab helps to perform different operations on a vast set of data like analysis through
what-if analysis tools and many other data analysis tools, removing duplicate data, transpose
the row and column, etc. It also helps to access data(s) from different sources as well, such as
from Ms-Access, from web, etc.

Review
This tab provides the facility of thesaurus, checking spellings, translating the text, and helps to
protect and share the worksheet and workbook.

View
It contains the commands to manage the view of the workbook, show/hide ruler, gridlines, etc,
freezing panes, and adding macros.

Department of Mechanical Engineering, DYPCET, Kolhapur.

67
 AIML for Mechanical Engineers

Advanced Functions and Formulas in Excel for Data Analysis:

Excel offers a wide range of advanced functions and formulas that are invaluable for
data analysis tasks. These functions enable users to manipulate, summarize, and analyze large
datasets efficiently. Here's a short overview:

VLOOKUP and HLOOKUP:

VLOOKUP and HLOOKUP functions are used to search for a value in a specified table
or range and return a corresponding value from a specified column or row.
They are commonly used for data lookup and retrieval tasks, such as retrieving sales figures
based on product codes or customer IDs.

INDEX and MATCH:

INDEX and MATCH functions are a powerful combination for performing advanced
lookup operations in Excel. INDEX returns the value of a cell in a specified row and column
of a table or range, while MATCH returns the relative position of a value in a specified range.
Together, INDEX and MATCH offer more flexibility and robustness compared to VLOOKUP
and HLOOKUP, especially for complex lookup scenarios.

SUMIFS, COUNTIFS, and AVERAGEIFS:

SUMIFS, COUNTIFS, and AVERAGEIFS functions allow users to perform
conditional sum, count, and average calculations based on multiple criteria.
They are useful for summarizing data based on specific conditions, such as calculating total
sales for a particular product category in a given month.

IF, IFERROR, and Nested IF:

IF function evaluates a specified condition and returns one value if the condition is true
and another value if the condition is false. IFERROR function returns a specified value if a
formula results in an error. Nested IF function allows users to perform multiple conditional
evaluations within a single formula.

TEXT and DATE Functions:

TEXT function converts numeric values, dates, and times into formatted text strings
based on specified format codes. DATE function returns the serial number of a date based on
the specified year, month, and day values. These functions are useful for manipulating date and
time data and formatting it according to specific requirements.

ARRAY Formulas:

Array formulas enable users to perform calculations on arrays of data rather than individual
cells. They are particularly useful for performing complex calculations, such as matrix
operations, statistical analysis, and multi-criteria calculations.
Some other functions useful in data analysis in Excel are Concat, Substitute, Trim, Filter etc.

Department of Mechanical Engineering, DYPCET, Kolhapur.

68
 AIML for Mechanical Engineers

Dashboards in Excel: Enhancing Data Visualization

Dashboards in Excel are powerful tools for visually summarizing and presenting key insights
from data in a concise and interactive format. They allow users to monitor performance, track
trends, and make data-driven decisions more effectively. Here's an overview of dashboards in
Excel:

Introduction:

A dashboard is a visual representation of data that provides a snapshot view of key performance
indicators (KPIs), metrics, and trends.
Excel dashboards typically consist of charts, graphs, tables, and interactive elements arranged
on a single worksheet or across multiple sheets.

Key Components of Excel Dashboards:

Charts and Graphs:

Dashboards often include various types of charts and graphs, such as bar charts, line graphs,
pie charts, and scatter plots, to visualize trends and patterns in data.
Users can choose the appropriate chart type based on the nature of the data and the insights
they want to convey.

Department of Mechanical Engineering, DYPCET, Kolhapur.

69
 AIML for Mechanical Engineers

Tables and PivotTables:

Tables and PivotTables are used to display detailed data summaries and allow users to drill
down into specific information.
PivotTables provide dynamic data summarization and filtering capabilities, making them
particularly useful for analyzing large datasets.

Interactive Controls:
Excel dashboards can incorporate interactive elements such as drop-down lists, scroll bars, and
option buttons to enable users to filter and explore data dynamically.
These controls enhance user engagement and allow for more personalized data analysis.

Conditional Formatting:
Conditional formatting is used to highlight important data points and trends visually.
Users can apply formatting rules based on specified criteria, such as color scales, data bars, and
icon sets, to draw attention to significant changes or outliers in the data.

Dashboard Layout and Design:

The layout and design of an Excel dashboard play a crucial role in its effectiveness and user
experience.
Clear organization, logical flow, and consistent formatting help users navigate the dashboard
efficiently and understand the information presented.
Data summarization and visualization are essential processes in data analysis that involve
condensing large datasets into meaningful insights and presenting them in a visually intuitive
manner. Here's a brief overview:

Excel Data Modelling and Automation Techniques

Data Modelling in Excel:

Excel data modelling involves structuring and organizing data within Excel to
facilitate analysis, reporting, and decision-making.
Techniques such as creating tables, defining relationships between data sets, and using
functions and formulas help in building data models.
PivotTables and PivotCharts are powerful tools in Excel for summarizing and analyzing data
models, providing dynamic views of the underlying data.
Automation Techniques in Excel:

Excel offers various automation techniques to streamline repetitive tasks and increase
productivity.

Macros: Macros are recorded sequences of actions that automate repetitive tasks in Excel.
They can be created using the Macro Recorder or written in VBA (Visual Basic for
Applications) code.

Formulas and Functions: Excel functions such as VLOOKUP, INDEX-MATCH, SUMIFS,

and IF statements automate calculations and data manipulation tasks, reducing manual effort.
Conditional Formatting: Conditional formatting automatically formats cells based on
predefined conditions, making it easy to visually identify trends, outliers, or exceptions in
data.

Department of Mechanical Engineering, DYPCET, Kolhapur.

70
 AIML for Mechanical Engineers

Data Validation: Data validation rules restrict the type and format of data that users can
input into cells, ensuring data accuracy and consistency.

Power Query: Power Query is a data connection and transformation tool in Excel that
automates the process of importing, cleaning, and shaping data from various sources.

Power Pivot: Power Pivot is an Excel add-in that enhances data analysis capabilities by
enabling users to create data models, perform advanced calculations, and generate interactive
reports.

Benefits:
Efficiency: Data modelling and automation techniques in Excel streamline workflows,
reduce manual effort, and increase productivity.
Accuracy: By automating calculations and data manipulation tasks, Excel helps minimize
errors and ensure data accuracy.
Consistency: Automation techniques ensure consistency in data formatting, calculations, and
reporting, leading to standardized and reliable outputs.
Scalability: Excel's data modelling and automation capabilities can scale to handle large
datasets and complex analysis requirements, making it suitable for both small-scale and
enterprise-level applications.

Department of Mechanical Engineering, DYPCET, Kolhapur.

71
 AIML for Mechanical Engineers

Best Practices for Creating Excel Dashboards:

Define Objectives:
Clearly define the purpose and objectives of the dashboard to ensure it meets the needs of its
intended audience.
Choose Relevant Metrics:
Select key metrics and KPIs that align with the objectives of the dashboard and provide
actionable insights.
Keep it Simple:
Avoid clutter and information overload by focusing on essential data points and using clear,
concise visuals.
Use Consistent Formatting:
Maintain consistency in formatting, color schemes, and font styles to enhance readability and
visual appeal.
Test and Iterate:
Test the dashboard with end-users and gather feedback to identify areas for improvement.
Iterate on the design based on user input and evolving business requirements.

Applications of Excel Dashboards:

Excel dashboards find applications across various industries and functions, including:
Sales and Marketing: Tracking sales performance, campaign effectiveness, and customer
engagement metrics.
Finance and Accounting: Monitoring financial metrics, budget vs. actuals, and cash flow
analysis.
Operations and Supply Chain: Managing inventory levels, production schedules, and
supplier performance.
Human Resources: Analyzing workforce demographics, employee turnover rates, and
performance metrics.

Dashboard Creation Process

Creating dashboards in Excel involves several steps to design, populate, and format the
dashboard effectively. Below is a detailed, step-by-step guide to creating Excel dashboards:

Step 1: Define Objectives and Audience

Identify the purpose of the dashboard and define the key objectives it should achieve.
Determine the target audience for the dashboard and their specific informational needs.

Step 2: Plan Dashboard Layout

Sketch a rough layout of the dashboard, including the placement of charts, tables, and other
elements.
Consider the flow of information and prioritize the most critical data points.

Step 3: Gather and Prepare Data

Collect the data required for the dashboard from relevant sources, such as Excel spreadsheets,
databases, or external files. Clean and organize the data to ensure accuracy and consistency.
Perform any necessary calculations or data transformations to derive the metrics and KPIs
needed for the dashboard.

Department of Mechanical Engineering, DYPCET, Kolhapur.

72
 AIML for Mechanical Engineers

Step 4: Choose Visualizations

Select appropriate chart types and visualizations based on the nature of the data and the
insights you want to convey. Consider using a mix of charts, graphs, and tables to present
different aspects of the data effectively. Choose visually appealing colors and styles to enhance
readability and engagement.

Step 5: Insert Charts and Graphs

Insert the selected charts and graphs onto the dashboard worksheet.
Use Excel's charting tools to create the desired visualizations, such as bar charts, line graphs,
pie charts, and scatter plots. Customize each chart's properties, including titles, axes, data
labels, and legend, to improve clarity and interpretation.

Step 6: Add Tables and PivotTables

Insert tables and PivotTables to display detailed data summaries and allow for further
exploration. Use PivotTables to summarize and analyze large datasets dynamically, enabling
users to filter and drill down into specific information.

Step 7: Incorporate Interactive Controls

Add interactive elements such as drop-down lists, scroll bars, and option buttons to
enable user interaction and data filtering. Link these controls to relevant data ranges or
PivotTable filters to update the dashboard dynamically based on user selections.

Step 8: Apply Conditional Formatting

Use conditional formatting to highlight important data points and trends visually. Apply
formatting rules based on specified criteria, such as color scales, data bars, and icon sets, to
draw attention to significant changes or outliers in the data.

Step 9: Arrange and Format Dashboard Components

Arrange the charts, tables, and interactive controls on the dashboard worksheet
according to the planned layout. Ensure that the dashboard is visually balanced and organized,
with clear labels and headings for each component. Format the dashboard elements
consistently, using a cohesive color scheme and font style to enhance readability and visual
appeal.

Step 10: Test and Refine

Test the dashboard with end-users to ensure it meets their informational needs and
objectives. Gather feedback and iterate on the design based on user input and evolving business
requirements. Make any necessary adjustments to the layout, content, or formatting to improve
usability and effectiveness.

Step 11: Finalize and Distribute

Finalize the dashboard design and ensure that all data and visualizations are up to date.
Save the dashboard as a separate Excel file or worksheet, making sure to protect sensitive data
if necessary. Distribute the dashboard to stakeholders or end-users, providing instructions on
how to use interactive features and interpret the displayed information.
By following these step-by-step instructions, you can create a well-designed and
functional dashboard in Excel that effectively communicates key insights and supports data-
driven decision-making.

Department of Mechanical Engineering, DYPCET, Kolhapur.

73
 AIML for Mechanical Engineers

Dashboards vs Interactive Dashboards

Aspect Dashboards Interactive Dashboards

Visual representation of data Dynamic and user-driven interfaces enabling
Definition providing snapshot views of real-time data exploration.
KPIs, metrics, and trends.
Typically display static Offer advanced features such as filtering,
Functionality information in the form of charts, sorting, drill-down, and data exploration
graphs, and tables. capabilities.
Lack interactive features or user Allow users to interact with data by selecting
controls. Users cannot manipulate specific data points, applying filters, and
Interactivity
or interact with the data. dynamically adjusting the displayed
information.
Provide a summary overview of Empower users to perform ad-hoc analysis,
data, allowing users to monitor identify insights, and gain deeper
Purpose
performance and track trends. understanding of the data by manipulating and
exploring different aspects of the dataset.
Sales dashboard displaying Sales performance dashboard with filter
monthly revenue trends, financial options to view data by region, product, or
Examples dashboard showing budget vs. time period, financial dashboard allowing
actuals. users to drill down into specific cost centers
or departments.

Department of Mechanical Engineering, DYPCET, Kolhapur.

74