AI PROJECT CYCLE: - process of converting real life problems into AI based models.

Project Cycle is a step-by-step process to solve problems using proven scientific methods and drawing
inferences about them.
Creating birthday card
 consider budget
 make list of gather data
 create model on collected data
 Show to someone else to evaluate it.

Components/ Elements/ Stages of AI project cycle:

PROBLEM SCOPING: - defines the process of selecting a problem which we might want to solve using AI
knowledge.
Identifying problem & then having a vision to solve is called problem scoping.
 understands prob
 finding factors which affects prob
 define goals/aim of project
4 W’s Canva of problem scoping
WHO: - directly/indirectly affected by problem, who will be benefitted from solution (known as
stakeholders)
WHAT: - identify and understand the nature of problem & gather evidence for existence of problem.
WHERE: - where problem arise? (Situation/ location/ context)
WHY: - is given problem worth solving? It thinks about the benefits of solution to stakeholders &
society
Problem solving template: - Summarizes key points of 4Ws canva into single template
The Stakeholders Who
Have problem Problem What
While/when Situation Where
Ideal solution Solution Why

ITERATIVE NATURE OF PROBLEM SCOPING: - Iterative approach continuously improves AI model.

It involves creating a prototype & testing it (repeating this cycle until you reach desired AI model)
Advantages of this approach: -
 Continuously improve design/product using AI model
 Cost – effective as continuous testing gives clear status of project
 Testing & debugging are easier with smaller iterations.
# debugging (process of identifying and removing errors from computer hardware/software)
 Can present result of each iteration to client/stakeholder to showcase the efficient progression
of project.
AI ETHICS
 Model should be understandable by all
 Every aspect should be self-explanatory & transparent
 Covering all sections of population
 Countercheck & validate results and all assumptions taken.
 Timely detection of errors.
DATA ACQUISITION: - collecting accurate & reliable data to work with.
It is a time – consuming process because variety of data is scattered everywhere and we need to focus on
data relevant to our needs.
DATA: - piece of raw information/ facts & statistics collected together for reference and analysis.
Data needs to be processed to interpret meaningful information out of it.
TYPES OF DATA SETS: -
TRAINING DATA TESTING DATA
Use is used to train the AI model Is used to test the AI model after its training
Size Huge amount of data (70 – 80%) Small amount of data (20 – 30%)
The efficiency of the AI system is dependent on the authenticity & relevance of the training data.
SYSTEM MAPS: -
 Diagrammatic representation of set of things working together.
 It helps us to find the relationships among different elements & find solution to achieve the
goal of our project.
 Used to understand complex issues that have interconnected factors affecting each other
 System maps comprises of:-
 Elements: - different – different elements within system
 Interconnections: - relationships that connects elements.
RULES FOR SYSTEM MAPS:-
 The circles represent elements.
 Arrows are used to represent relationships/interconnections.
 The + and - signs are indicators of the nature of a relationship. The arrowhead depicts the
direction of the effect and the sign (+ or -) shows their relationship.
 If the arrow goes from X to Y with a + sign, it means that both are directly related to each
other. That means if X increases, Y also increases and vice versa.
 If the arrow goes from X to Y with a -sign, it means that both the elements are inversely related
to each other That means if X increases, Y would decrease and vice versa.
DATA ACQUISITION TECHNIQUES:-
 WEB SCRAPING: - collecting data from web using technologies(online data) (e.g.:- monitoring
prices)
 SENSORS: - collect physical data and detect changes(live data in offline mode)
 CAMERAS: - capture visual information(image) used as a source of data
 API: - Application Programming Interface, it’s a messenger that takes request & tells system
what to process that request then respond accordingly (e.g.:- Google API, Twitter API)
 OBSERVATIONS: - collects data by watching facts as they occur. This data is generally used to
test model. It’s a time consuming data source.
 SURVEYS: - gather specific information from sample of people (e.g.:-census survey for
analyzing population through questionnaire). Surveys can be online, telephonic, in person)

DATA EXPLORATION: - interpreting/finding useful information (patterns & trends) out from data acquired.
 Explore data
 Arrange it uniformly
 Validate/verify data
 Is data gathered according to requirement (specifications decided)?
 Is gathered data free from error?
 Is it meeting our needs?
 DATA CLEANING: - getting rid of commonly found errors & mistakes in data set such as:
 Outliers: data points existing out of range/ distant from other observations
 Missing data
 Erroneous data: incorrect data points/ rejected
DATA VISUALISATION: - graphical representation of data & information
NEED FOR DATA VISUALISATION: -
 Graphical representation makes data easier to understand and communicate which gives
sense of trends, patterns, relationships and reveal hidden patterns also.
 Helps us to make sense out of huge data.
 Helps in strategy building
DATA VISUALISATION TECHNIQUES: -
 LINE GRAPH: - graphs that uses lines to connect individual data points. It is most frequently
used for quantitative values over continuous intervals (e.g.: - daily profit earned by
shopkeeper)
 AREA GRAPH: - display development of quantitative values over intervals. It is extension of
line chart. (e.g.: - showing trends over time by comparing them: - most preferred show to
watch in Netflix)
 BAR CHART: - horizontal/vertical, display comparisons among non – continuous development
over time (e.g.: - rise in population in 5yrs)
 HISTOGRAM: - continuous development over time.
 PIE CHART: - display circular representation of data that shows proportions and percentages
between categories by dividing the circle (100%) into segments (each showing relative size of
data). This gives quick idea of distribution of data. (e.g.: - marks of all students in class)
 BUBBLE CHART: - combination of bubble chart + data visualization + map. It visualize location
and proportions using circles over geographical regions with area of circle proportional to
value in data set (e.g.: - no. of tigers in national park in India)

DATA MODELLING: -

RULE - BASED

AI MODELS
MACHINE LEARNING

LEARNING BASED
DEEP LEARNING

RULE BASED APPROACH: - based on rules and facts defined by developer, fed to machine to perform task
and generate output accordingly. So, it’s a static model i.e. the machine once trained, does not take into
consideration any changes made in the original training dataset. (e.g.: - weather forecast AI model)
LEARNING BASED APPROACH: - dynamic approach AI modelling where the machine learns by itself.
Relationships or patterns are not defined by the developer. Random data is fed into machine and machine
develops its own patterns or trends based on data outputs. The machine tries to extract similar features &
then cluster them into same datasets.

ARTIFICIAL INTELLIGENCE MACHINE LEARNING DEEP LEARNING

Mimics human intelligence Can learn through data and solve Helps in discovering patterns & trends.
complex problems
 Subset of data science Subset of AI Subset of machine learning

It is the stimulation of It is the training of machines to It is the process of using artificial neural
intelligence in machines. take decisions with experience. network for solving complex problems

DECISION TREE – RULE BASED APPROACH: - decision trees are tools that follow rule – based approach. It’s
a kind of flowchart where flow starts at root node & ends with a decision made at leaves. It is used to depict
conditions and their outcomes.
 Root node: - 1st node, represents entire set of data with two different ways/condition (yes/no)
 Branching: - dividing node at one level into 2 or more sub nodes at next level. The forks or diversions
are known as branches of tree.
 Decision node: - dividing a node further into another level sub – node.
 Leaf node: - A node that does not split further.
 Parent node: - a node that is level above a sub – node.
 Child node: - a sub – node that falls under another node.
Following are some of the important points to consider while designing a decision tree.
 Possibility of multiple decision trees which lead to correct prediction for a single dataset. The simplest
one should be chosen.
 Dataset might contain redundant data at times, so it is necessary that only those parameters that
affect the output directly should be included.
 Try selecting any one output and on its basis, find out the common links which all the similar outputs
have.

AI PROJECT EVALUATION: - Evaluation is a process of understanding the reliability of any AI model, based
on outputs by feeding the test dataset into the model and comparing it with actual answers. When AI model
is evaluated for its efficiency and accuracy, it enables continuous improvement of the model to achieve the
project goals. The model must be tested with varied data to ensure that the results are satisfactory. Once
the model is evaluated it must be deployed.
The efficiency of the model is calculated on the basis of the parameters mentioned below:
 Accuracy: - percentage of correct predictions out of all the observations.
 Precision: - percentage of true positive cases versus all the cases where theprediction is true.
 Recall: - fraction of positive cases that are correctly Identified.
 F1 Score: - number between 0 and 1 and is the harmonic mean of precision and recall.

AI PROJECT DEPLOYMENT: - Deployment is the process of integrating a newly created Al model into an
existing production environment to make practical implementation of the model with actual data taken as
input to give the desired output. It requires certain settings to be done in terms of hardware and software
so that the Al model can be put to use efficiently by the end users.