Module 1

1. Overview & Definition of Analytics

• Analytics refers to the use of data, statistical analysis, and modeling to solve problems, gain
insights, and make decisions.

• It involves collecting, processing, and interpreting data to find meaningful patterns.

2. Need for Analytics

• Organizations generate huge amounts of data; analytics helps convert this data into
actionable insights.

• Helps improve efficiency, reduce costs, and predict trends.

• Necessary for staying competitive in today’s data-driven world.

3. Analytics in Decision Making

• Analytics supports evidence-based decision making.

• Reduces guesswork and increases the accuracy of business decisions.

• Used in areas like marketing, finance, operations, and HR to make better decisions.

4. Analytics as a Game Changer and Innovator

• Transforms how businesses operate by providing real-time insights.

• Drives innovation by identifying new opportunities.

• Examples: Personalized recommendations by Amazon or Netflix.

5. Power of Analytics

• Enables data-driven strategies.

• Helps in understanding customer behavior, market trends, and operational performance.

• Supports automation and optimization of processes.

6. Predictive Analytics

• A branch of analytics that uses historical data and statistical models to predict future
outcomes.

• Common tools: Regression analysis, machine learning, forecasting models.

• Example: Predicting customer churn or future sales.

 Module 2

1. Types of Predictive Analytics

• These are the main approaches to predicting future outcomes:

• Classification Models
➤ Predict categories (e.g., spam or not spam)
➤ Used in email filters, fraud detection

• Regression Models
➤ Predict numeric values (e.g., sales, revenue)
➤ Used in forecasting

• Time Series Analysis

➤ Predict future trends based on time-based data
➤ Used in stock price or demand forecasting

• Clustering
➤ Group similar data (not direct prediction but used for segmentation)
➤ Used in market research

• 2. Techniques of Predictive Analytics

• These are the tools and methods used:

• Linear & Logistic Regression

➤ Predict continuous and binary outcomes

• Decision Trees
➤ Easy-to-understand visual models for decision-making

• Neural Networks & Deep Learning

➤ Powerful for complex patterns like image/speech recognition

• Random Forest, XGBoost

➤ Advanced ensemble techniques for higher accuracy

• Machine Learning (ML)

➤ Used for self-learning predictive models

• 3. Applications of Predictive Analytics

• Manufacturing

• Healthcare

• Personalized treatment plans

• Telecommunication
• Network optimization

• Fraud detection

• Supply Chain

• Inventory forecasting

• Optimize delivery routes

• Supplier risk assessment

• Information Technology

• Cyber threat prediction

• Server failure prediction

• IT resource planning

• 4. Digital Analytics (Simplified)

• Definition:

• Digital Analytics is the analysis of digital data from websites, mobile apps, social media, etc.,
to optimize user experience and business outcomes.

• Key Focus Areas:

• Website traffic (page views, bounce rate)

• User behavior (clicks, time spent)

• Conversion tracking (sales, signups)

• Social media metrics (likes, shares, comments)

• Tools Used:

• Google Analytics

• Adobe Analytics

• Social media insights (Meta, X, LinkedIn)

• Heatmaps (Hotjar, Crazy Egg)

MODULE 3

• Simple Linear Regression (SLR) is a statistical method to predict the value of one variable (Y)
using one independent variable (X).

• It fits a straight line to the data:

Y = β₀ + β₁X + ε
Where:

o Y = dependent variable (outcome)

 o X = independent variable (predictor)

o β₀ = intercept

o β₁ = slope

o ε = error term

2. Importance of SLR

• Helps in predicting outcomes (e.g., predicting sales based on advertising spend)

• Useful for understanding relationships between variables

• Forms the foundation for more advanced regression models

3. Types of Regression (Basic Mention)

• Simple Linear Regression – one independent variable

• Multiple Linear Regression – more than one independent variable

4. SLR Model Building Steps

1. Identify variables – Define X and Y

2. Collect data – Numerical, continuous values

3. Plot the data – Scatter plot to check linearity

4. Fit the line – Use regression to find β₀ and β₁

5. Evaluate model – Check goodness of fit

5. OLS Estimation (Ordinary Least Squares)

• A method to find best-fitting line by minimizing the sum of squared errors (differences
between actual and predicted Y)

• Gives estimates of β₀ (intercept) and β₁ (slope)

6. Model Interpretation

• Intercept (β₀): Value of Y when X = 0

• Slope (β₁): Change in Y for a one-unit change in X

• R² (R-squared): Explains how well the model fits the data (ranges from 0 to 1)
 7. Model Validation

• Residual analysis: Errors should be randomly scattered (no pattern)

• Check assumptions:
Linearity
Independence
Homoscedasticity (constant variance)
Normality of residuals

• Use metrics:

o R²

o RMSE (Root Mean Square Error)

o p-value (for statistical significance of β₁)

MODULE 4

1. Multiple Linear Regression (MLR): Introduction

• MLR is used to predict the value of a dependent variable (Y) using two or more
independent variables (X₁, X₂, ..., Xn).

• General equation:
Y = β₀ + β₁X₁ + β₂X₂ + ... + βnXn + ε
Where:

o Y = Dependent variable

o X₁, X₂, ... Xn = Independent variables

o β₀ = Intercept, β₁...βn = Coefficients

o ε = Error term

2. Estimation of Regression Parameters

• Done using Ordinary Least Squares (OLS) method

• Objective: Minimize sum of squared residuals (errors)

• Output: Coefficients (β-values) that best fit the data

3. Model Diagnostics

Helps check if the model is valid and reliable:

• R²: How well independent variables explain variation in Y

• Adjusted R²: Better for MLR; adjusts for number of variables

• Residual plots: Should be randomly scattered

• p-values: Check if variables are statistically significant (p < 0.05)

• F-test: Tests overall significance of the model

4. Dummy, Derived & Interaction Variables

Dummy Variables

• Used to represent categorical data (e.g., Male = 0, Female = 1)

• Needed because regression requires numeric input

Derived Variables

• Created by transforming existing variables

• Example: log(salary), age²

Interaction Variables

• Show combined effect of two variables

• Example: X₁ * X₂ — effect of X₁ depends on X₂

5. Multicollinearity

• When independent variables are highly correlated with each other

• It distorts regression results

• Detected using:

o VIF (Variance Inflation Factor): VIF > 10 indicates a problem

• Solution:

o Drop one of the correlated variables

o Use Principal Component Analysis (PCA)

6. Model Deployment

• Deploying a model means using it in real-world applications

• Steps:

1. Finalize model

2. Convert into software/script

3. Integrate into systems (e.g., websites, apps)

4. Monitor performance
 7. Demo Using Software

• Tools like Excel, Python, R, SPSS, or SAS used for MLR

• Common steps in software:

➤ Import data → Fit model → Interpret output (coefficients, R², p-values) → Plot diagnostics

MODULE 5

Logistic regression is a supervised machine learning algorithm used for classification tasks where
the goal is to predict the probability that an instance belongs to a given class or not. Logistic
regression is a statistical algorithm which analyze the relationship between two data factors. The
article explores the fundamentals of logistic regression, it's types and implementations.

• It can be either Yes or No, 0 or 1, true or False, etc. but instead of giving the exact value as
0 and 1, it gives the probabilistic values which lie between 0 and 1.

• In Logistic regression, instead of fitting a regression line, we fit an "S" shaped logistic
function, which predicts two maximum values (0 or 1).

MODULE 6

1. Overview of Decision Trees

• A decision tree is a flowchart-like model used for classification and regression tasks.

• It splits data into branches based on conditions to make predictions.

• Simple to visualize, interpret, and explain.

2. Applications of Decision Trees

Used in many fields like:

• Marketing → Predict customer behavior

• Finance → Credit risk scoring

• Healthcare → Disease diagnosis

• HR → Predict employee attrition

5. Introduction to CHAID

• CHAID = Chi-Square Automatic Interaction Detector

• Used for categorical target variables

• Splits based on Chi-square tests (checks statistical significance)

• Creates multi-way splits (unlike binary splits in CART)

6. Classification and Regression Tree (CART)

• Common decision tree algorithm

 • Uses binary splits only (Yes/No)

• Two types:

o Classification Tree: For categorical output (e.g., yes/no)

o Regression Tree: For numeric output (e.g., price, salary)

• Splits based on:

o Gini Index (for classification)

o Mean Squared Error (MSE) (for regression)

MODULE 7
1. Introduction to Unstructured Data Analysis

What is Unstructured Data?

• Data that doesn’t follow a fixed format

• Examples:

o Text (emails, reviews, tweets)

o Images, videos, audio

Why is it important?

• 80–90% of data today is unstructured

• Analyzing it helps discover hidden insights (e.g., customer opinions, trends)

2. Sentiment Analysis

Definition:

• A technique to determine whether text expresses positive, negative, or neutral emotion.

Example:

• “This product is amazing!” → Positive

• “Worst service ever.” → Negative

Applications:

• Customer feedback analysis

• Social media monitoring

• Brand reputation tracking

 Techniques:

• Lexicon-based: Uses predefined word lists

• Machine Learning-based: Uses models like Naïve Bayes or SVM

3. Naïve Bayes Algorithm

Overview:

• A supervised machine learning algorithm based on Bayes' Theorem

• Called "naïve" because it assumes all features (words) are independent of each other

Formula:

• P(A|B) = [P(B|A) * P(A)] / P(B)

o A = class (e.g., positive/negative)

o B = input data (e.g., words in review)

How it works (for text):

1. Learn probabilities of words in each class

2. For new text, calculate probability for each class

3. Choose the class with the highest probability

Applications:

• Spam detection

• Email filtering

• Sentiment classification

• Text categorization

MODULE 8
1. Forecasting

• Forecasting is predicting future values based on past data.

• Used in sales, finance, inventory, weather prediction, etc.

• Helps in planning and decision making.

2. Time Series Analysis

• Deals with data collected over time at regular intervals (daily, monthly, yearly).

• Goal: Identify patterns/trends to forecast future points.

• Components of time series:

o Trend: Long-term upward or downward movement

o Seasonality: Repeating patterns at fixed periods (e.g., sales peak in December)

o Cyclic: Irregular, long-term fluctuations

o Random (noise): Unpredictable variations

3. Additive & Multiplicative Models

• Additive Model:
Time series = Trend + Seasonality + Noise
Use when seasonal variations are constant over time

• Multiplicative Model:
Time series = Trend × Seasonality × Noise
Use when seasonal variations increase/decrease proportionally with trend

4. Forecasting Accuracy

• Measures how close forecasted values are to actual values.

• Common metrics:

o MAD (Mean Absolute Deviation)

o MSE (Mean Squared Error)

o MAPE (Mean Absolute Percentage Error)

5. Moving Average Models

• Smoothes time series by averaging data points over a fixed window.

• Types:

o Simple Moving Average (SMA): Equal weights to all points

o Weighted Moving Average (WMA): Different weights, recent data given more
importance

6. Exponential Smoothing Techniques

• Give more weight to recent observations using a smoothing factor (α, 0 < α < 1).

• Types:

o Simple Exponential Smoothing: For data without trend or seasonality

 o Holt’s Linear Trend Model: For data with trend

o Holt-Winters Model: For data with trend and seasonality

Quick Exam Summary:

Topic Key Points

Forecasting Predict future using past data

Time Series Components Trend, Seasonality, Cyclic, Random

Additive vs Multiplicative Additive = add parts; Multiplicative = multiply parts

Accuracy Metrics MAD, MSE, MAPE

Moving Average Smooth data by averaging

Exponential Smoothing Recent data weighted more