Introduction to Business

Analytics
02/24/2025 1
Evaluation:

Students will be evaluated based on case analysis, group projects, class

participation, and quizzes. The weightage of each evaluation components
are listed below:

 Class participation: 10%

 Quizzes: 20%
 Project: 30%
 End Term Exam: 40%

24/02/2025 2
Topics:

 What is Analytics?
 Why Analytics?
 Types of Analytics

02/24/2025 3
 Analytics is the scientific process of transforming data
into insights for making better decisions.
OR

 Analytics is the process of discovering, interpreting, and

communicating significant patterns in data and using tools
to empower entire organization.

 Business analytics adds even more opportunities to drive

desired outcomes, such as optimization, cost savings, and
customer engagement.

02/24/2025 4
 Why Analytics ?
 Decision making is not easy, specially if the business is big
 Competitive Advantage
 Improving Efficiency

Example:

 City bank managers decide if a customer is trustworthy

 Amazon decides who is getting pregnant and hence offer on
baby care products
 Manager decides who to give proportional offers
02/24/2025 5
Give me some examples where industries are
making money through the use of data
analytics.
 1. Product Recommendations

 Product recommendation is one

of the most popular applications of
Analytics.

 It works as a filtering system that

predicts and shows items a user
might want to purchase.

 Product recommendation is a key

feature of almost every
e-commerce website.

 Analytics tracks user behavior

based on previous purchases,
search patterns, and cart history
to make these recommendations.

02/24/2025 7
 2. Image
 Image recognition is one of the most
Recognition
significant machine learning applications,
used to catalog and detect objects or
features in digital images.

 It helps identify places, logos, people,

objects, buildings, and many other
elements in images.

 This technology is widely adopted for

further analysis like pattern recognition,
face detection, and face recognition.

The focus is on how image recognition

enhances various industries by identifying and
processing visual data for deeper insights.

02/24/2025 8
 3. Sentiment
Analysis
 Sentiment analysis is a real-time machine
learning application that identifies the emotion
or opinion of the speaker or writer.

 For example, if someone writes a review, email,

or document, the sentiment analyzer detects
the thought and tone behind the text.

 It is commonly used in review-based websites

and decision-making applications to analyze
user feedback and other textual data.

This emphasizes how sentiment analysis helps in

understanding public opinion and making informed
decisions based on textual content.

02/24/2025 9
 4. Smart Health Records

 Maintaining health records is a time-consuming

process, even with technological advancements.

 Machine learning in healthcare aims to streamline

processes, saving time, effort, and money.

 MIT is at the forefront of developing smart health

records using machine learning for diagnosis and
clinical treatment suggestions.

This highlights the potential of machine learning to

transform healthcare by improving efficiency and
enhancing decision-making through intelligent health
records.

02/24/2025 10
 5. customer segmentation
 Customer segmentation involves dividing
customers into groups based on common
characteristics to enable effective and
targeted marketing.

 In business-to-business (B2B) marketing,

companies may segment customers by factors
such as industry and number of employees

 In B2C marketing, companies may segment

customers by factors such as Demographics,
Geographic's, Psychographics, Behavioral data
and Technographic.

 These segments help companies create

personalized marketing campaigns that resonate
with individual consumer needs and preferences.

02/24/2025 11
 6. Fraud detection
Machine learning is widely applied to detect
and prevent fraud across various industries.

In financial services, it identifies

fraudulent transactions, combats tax
evasion, and detects money laundering.

The insurance industry leverages

machine learning to uncover fraudulent
claims and activities

In retail banking, machine learning

enhances fraud prevention by analyzing
transaction patterns and anomalies.

This highlights machine learning's critical role in

proactively identifying and combating fraud in
different sectors.
02/24/2025 12
 7. Dynamic Pricing
 Dynamic pricing (also known as surge pricing,
demand pricing, or time-based pricing) is a flexible
pricing strategy that adjusts prices based on current
market demands.

 Machine learning algorithms analyze historical data

and integrate it with real-time market and consumer
insights to inform pricing decisions.

 This strategy enables businesses to optimize prices

based on numerous variables, ultimately helping to
maximize revenue.

 Examples of industries utilizing dynamic pricing

include:
1. Airline industry
2. Indian railway and transportation industry

02/24/2025 13
 Websites To Learn More About Data Science And
Machine Learning

https://www.kaggle.com/

https://archive.ics.uci.edu/ml/index.php

https://www.analyticsvidhya.com/

02/24/2025 14
 Type of
Analytics

02/24/2025 15
 Descriptive Analytics: What
happened?
Example of Descriptive Analytics:

 Monthly Profit and Loss Statement: Serves as a classic

example, summarizing financial performance over a specific
period.

 Customer Demographics: Demographic Insights:

Analyzing customer information, such as the fact that 30% of
customers are self-employed, falls under descriptive analytics
as it describes the customer base.

 Time-Based Analysis: Highlights the ability to compare data

across different time periods to observe changes in health
outcomes.

Purpose of Descriptive Analytics: Aims to provide a clear

view of past events and patterns, facilitating better decision-
making and strategic planning.

16
 Diagnostic Analytics: Why is it
happening?
Diagnostic analytics examines data to
answer the question, “Why did it
happen?”

 In the assessment of descriptive data,

diagnostic analytical tools empower
analysts to drill down and isolate the
root cause of a problem.

 For example, we can find the reason

why the recovery and death rates have
increased for COVID-19 patients in
April 2021 compared to April 2020.
17
 Predictive Analytics: What is
likely to happen?
Predictive analytics uses historical data and algorithms to
forecast future events or behaviors.

 Analyzes customer behavior and market trends to

create predictive models.
 Example: Retailers predict future buying patterns
based on past purchase data.

 Used in healthcare, finance, marketing, and more for

informed decision-making.
 Example: Hospitals forecast patient admission rates to
optimize resource allocation.

Note: Predictive analytics involves using historical data and

statistical algorithms to forecast future events or behaviors. It
analyzes patterns and trends to predict what is likely to occur.

18
 Prescriptive Analytics: What do I need to do?

Prescriptive analytics uses data and algorithms to recommend

actions for optimal outcomes.

 Data Utilization: Combines historical data, predictive

models, and business rules to guide decision-making.

 Techniques: Employs methods like optimization,

simulation, and decision analysis.

 Example: Airlines use it to optimize flight schedules based

on demand forecasts.

 Example: Supply chain managers use optimization models

to minimize costs while meeting demand.

 Example: Hospitals determine the best staffing levels

based on patient inflow predictions to enhance service
delivery.
19
02/24/2025 20
Longitudinal Data
 Subscription Age Salary Gender
Cross Sectional Data 0
0
25
64
132
84
1
1
0 31 100 0
0 66 72 1
0 39 78 1
A cross sectional data is collected by observing various 1 53 54 1
subjects like (firms, countries, regions, individuals), at the 1 49 102 0

same point in time or during the same period. 1 63 145 0

0 66 130 1
0 72 87 0
0 35 89 1
1. Sales of 30 companies 0 42 97 1
2. Productivity of each sales division 1 65 90 1
3. Gross annual income for each of 1000 randomly chosen 1 48 90 1
households in New York City for the year 2000 0 28 67 1
0 46 98 1
0 31 55 1
0 41 103 1
1 57 128 1
1 70 72 1

02/24/2025 34
 Time Series Data
Data collected for a single variable such as sales of smartphones over several time
interval (weekly, monthly) is called a time-series data.

Year Quarter Sales

2012 1 $1,65,000.00
2 $2,53,000.00
3 $3,16,000.00
4 $2,87,000.00
2013 1 $2,57,000.00
2 $3,08,000.00
3 $3,16,000.00
4 $3,51,000.00

02/24/2025 35
 Panel Data
A data collected for a several variables over several time interval
(weekly, monthly, yearly) is called Panel data.
Panel Data
Person Year Income Expenditure
$1,300.00
1 2016 $1,300.00
$1,300.00
1 2017 $1,600.00
$1,300.00
1 2018 $2,000.00
$1,300.00
2 2016 $2,000.00
$1,300.00
2 2017 $2,300.00
$1,300.00
2 2018 $2,400.00

02/24/2025 36
 Problem Statement
Project Title: Analyzing Factors Influencing Vehicle Prices

Step 1: Define the Problem

Objective of the Project: The core objective is to analyze vehicle advertisements and determine what
drives vehicle prices. This involves identifying key features or trends in the data that influence price
fluctuations and help predict vehicle valuation.

Research Questions:
• What vehicle attributes significantly affect the price?
• How do external factors (e.g., market trends, location) impact pricing?
• Can we predict vehicle prices accurately based on available data?
Step 2: Data Collection

Gather Historical Data: Collect advertisement data over the last few years, including:

• Vehicle make and model

• Year of manufacture
• Mileage
• Condition (new/used)
• Listing price
• Location of the advertisement
• Seller type (private seller, dealership)
• Date of the advertisement

Step 3: Data Preparation

Data Cleaning:
• Remove duplicates and irrelevant entries.
• Handle missing values appropriately (impute, drop, or fill).
• Standardize data formats (e.g., date formats, string casing).

Feature Engineering:
• Create new variables if necessary (e.g., age of the vehicle, price per mile).
• Convert categorical variables into numerical ones using techniques like one-hot encoding.
Step 4: Exploratory Data Analysis (EDA)

Visualizations:
• Use histograms, scatter plots, and box plots to visualize price distributions and relationships.
• Examine correlation matrices to identify relationships between features.

Statistical Analysis:
• Calculate descriptive statistics (mean, median, mode) for key features.
• Perform hypothesis testing to determine significant factors influencing price.

Step 5: Modeling

Select Appropriate Models:

• Start with simple regression models (e.g., linear regression) to establish baseline performance.
• Explore more complex models (e.g., decision trees, random forests, or gradient boosting) for improved accuracy.

Model Training and Testing:

• Split the dataset into training and testing sets (e.g., 80/20 split).
• Train models on the training set and evaluate performance on the testing set using metrics like RMSE, MAE, and R².
Step 6: Results Interpretation

Analyze Model Outputs:

• Determine which features are the most important in predicting vehicle prices.
• Visualize the relationship between predicted and actual prices.
•Generate Insights:
• Provide actionable insights based on analysis (e.g., optimal price ranges for specific vehicle types).

Step 7: Reporting

Create a Presentation:
• Summarize findings in a clear and concise manner.
• Use visual aids (charts, graphs) to illustrate key points.
•Recommendations:
• Suggest strategies for sellers based on the analysis.
• Propose improvements for Crankshaft List based on insights gained.

Step 8: Implementation

Integrate Findings:
• Work with relevant teams to implement findings into the platform.
• Monitor outcomes and adjust strategies based on ongoing analysis.
Data Reading
 Test: IF data distribution is
symmetric, replace with Mean

If data is numeric

Test: IF data distribution is not

symmetric, replace with
Median
Missing Value Imputation

Replace missing value with

If data is Non-numeric Mode
Imputation Technique Type of Data How it Works Best For

Replace missing values with the mean of

Mean Imputation Numerical the column. Normally distributed data.

Replace missing values with the median of Skewed numerical data (e.g., income, house
Median Imputation Numerical the column. prices).

Replace missing values with the most

Mode Imputation Categorical frequent value. Categorical features.

Forward/Backward Fill Time Series Propagate the next/previous value. Sequential or time-series data.

Both Numerical & Use nearest neighbors to impute missing Complex datasets with relationships between
KNN Imputation Categorical values. features.

MICE (Multivariate Interdependent missing data across multiple

Imputation) Both Iteratively predicts missing values. features.

Randomly select a value from existing

Random Sampling Both data. Preserving data distribution.

Use regression or other models to predict Well-understood relationships between

Predictive Models Both values. variables.

Estimate missing values through

Interpolation Time Series linear/curve fitting. Continuous time-series data.

When missing data is too extensive or

Drop Missing Values Both Remove rows/columns with missing data. irrelevant.
Numeric Dataset
Outlier detection
 Why might advertisement dates such as week, month,
and year be useful in predicting the price of an old
vehicle?

How can we impute the average distance driven,

which is an important factor in the valuation of
used vehicles?
Histogram
Outlier Detection
Pie Chart
Question: Study how many days advertisements were displayed (days_listed). Plot a histogram, calculate
the mean and median, and describe the typical lifetime of an ad. Identify when ads were removed quickly
and when they were listed for an abnormally long time.
Continue…..
ANOVA
ANOVA Continue…