UNIT -4

PART 1- PREDICTIVE ANALYTICS & MODELING

PART2 - DATA REDUCTION TECHNIQUES

PART 1- PREDICTIVE MODELING & ANALYSIS

Predictive modeling involves finding good subsets of predictors or explanatory variables. Models
that fit the data well are better than models that fit the data poorly. Simple models are better
than complex models. Working with a list of useful predictors, we can fit many models to the
available data, then evaluate those models by their simplicity and by how well they fit the data.

Predictive modeling is a data-driven technique used in business analytics to forecast future

outcomes based on historical data and statistical algorithms. It involves identifying patterns,
relationships, and trends in data to make predictions and informed decisions.

Key Features of Predictive Modeling and Analysis:

Historical Data Utilization: Predictive models rely on historical data to identify patterns and
trends. This data can be collected from various sources, including customer records, sales
transactions, or website interactions.

Data Preprocessing: Before modeling, data must be cleaned, transformed, and standardized
to ensure accuracy and consistency. This includes handling missing values, outlier detection,
and feature engineering.

Target Variable: Predictive modeling centers around a target variable, the outcome we want to
predict. It could be binary (yes/no), categorical (e.g., customer segments), or continuous (e.g.,
sales revenue).

Independent Variables (Features): These are the variables used to make predictions. They
can be quantitative or qualitative and are selected based on their potential to influence the
target variable.

Algorithm Selection: Choosing the right predictive algorithm is crucial. Common algorithms
include linear regression, decision trees, logistic regression, and machine learning techniques
like Random Forest, Gradient Boosting, or Neural Networks.

Model Training: The model is trained on a portion of the historical data, learning the
relationships between the independent and target variables.

Validation and Testing:Models need to be validated and tested to ensure they perform well.
This involves splitting the data into training and testing sets to evaluate the model's accuracy,
precision, recall, and other metrics.
Cross-Validation: To minimize overfitting and assess model generalizability, cross-validation
techniques like k-fold cross-validation are used.

Feature Importance:Identifying which independent variables have the most impact on the
target variable is crucial for understanding the model's insights and business decisions.

Model Deployment:Once a model is validated and ready, it can be deployed for real-world
predictions and decision-making, often integrated into business processes.

Model Interpretability:Understanding the factors and reasoning behind predictions is essential

for building trust and making actionable decisions.

Continuous Monitoring:Predictive models require ongoing monitoring and maintenance to

adapt to changing data patterns and ensure their accuracy remains high.

Business Impact:The ultimate goal of predictive modeling is to generate business value, such
as increased revenue, cost reduction, improved customer retention, or enhanced
decision-making.

Three general approaches to research and modeling as employed in predictive analytics:

traditional, data-adaptive, and model-dependent.

The traditional approach to research and modeling begins with the specification of a theory or
model. Classical or Bayesian methods of statistical inference are employed. Traditional
methods, such as linear regression and logistic regression, estimate parameters for linear
predictors. Model building involves fitting models to data. After we have fit a model, we can
check it using model diagnostics.

When we employ a data-adaptive approach, we begin with data and search through those
data to find useful predictors. We give little thought to theories or hypotheses prior to running the
analysis. This is the world of machine learning, sometimes called statistical learning or data
mining. Data adaptive methods adapt to the available data, representing nonlinear relationships
and interactions among variables.

Model-dependent research is the third approach. It begins with the specification of a model
and uses that model to generate data, predictions, or recommendations. Simulations and
mathematical programming methods, primary tools of operations research, are examples of
model-dependent research.

LOGIC DRIVEN & DATA DRIVEN MODELING

Logic Driven Modeling

Logic Driven Modeling is an approach to business analytics that relies on predefined business
rules and expert knowledge to make decisions and predictions.
It is based on formal logic, which uses if-then rules to infer conclusions.
Logic Driven Modeling is often used in situations where the decision-making process is
well-understood and can be codified.

Features:
Rule-Based: Logic Driven Modeling relies on predefined rules or conditions that dictate how
decisions are made.
Expert Knowledge: It incorporates domain expertise and the collective knowledge of subject
matter experts.
Transparency: The decision-making process is transparent, as it is based on explicit rules and
logic.
Deterministic: The outcomes are predictable and consistent since they follow predefined rules.
Interpretability: It is easy to understand and interpret the reasoning behind decisions, making it
useful for compliance and regulatory requirements.

Applications:
● Logic Driven Modeling is commonly used in credit scoring, fraud detection, and
compliance analysis.
● It is suitable for scenarios where there are well-defined business rules and regulatory
requirements.

Challenges:

● Limited Flexibility: Logic Driven Models may not adapt well to changing conditions or
dynamic environments.
● Requires Expert Input: Creating and maintaining rules demands domain expertise and
constant rule updates.

Data Driven Modeling

Data Driven Modeling is an approach that uses historical data to make predictions and
decisions, often without explicitly defined rules. It leverages statistical and machine learning
techniques to discover patterns and relationships in data.

Features:

● Data-Centric: Data Driven Modeling focuses on the data and the insights it can provide.
● Adaptability: It can adapt to changing data and evolving conditions, making it suitable
for dynamic environments.
● Complexity Handling: It can handle complex, non-linear relationships in data.
● Predictive Power: Data Driven Models can provide highly accurate predictions based
on historical data.
Applications:
● Data Driven Modeling is widely used in recommendation systems, predictive
maintenance, and customer churn analysis.
● It is suitable for scenarios where data is abundant and the decision-making process is
not explicitly defined.

Challenges:

● Black Box: Data Driven Models can be challenging to interpret and may lack
transparency, especially in complex models.
● Data Quality: The accuracy of predictions depends on the quality and quantity of
historical data.
● Overfitting: Data Driven Models can overfit to noise in the data, leading to poor
generalization.

STRATEGIES FOR BUILDING PREDICTIVE MODELS

1. Data Collection and Preparation:

● The first step in building predictive models is to gather high-quality data from reliable
sources.

● Ensure data cleaning, which involves handling missing values, outliers, and
inconsistencies.
● Transform and preprocess the data by encoding categorical variables and scaling
numerical ones to make it suitable for modeling.

2. Define the Problem and Objectives:

● Clearly articulate the problem you want to solve and define your objectives. What do you
aim to predict or optimize?

● Identify the relevant variables and the target variable (what you want to predict).

3. Exploratory Data Analysis (EDA):

● Conduct EDA to gain insights into the data. Use visualization and summary statistics to
understand data patterns.

● Identify potential relationships, trends, and correlations that can inform your modeling
approach.

4. Feature Selection and Engineering:

● Choose the most relevant features (variables) for your predictive model.
 ● Create new features that may capture hidden patterns or relationships in the data.
● Use techniques like feature importance ranking and dimensionality reduction.

5. Model Selection:

● Choose an appropriate modeling technique based on the nature of the problem

(classification, regression, clustering, etc.).

● Consider the strengths and weaknesses of algorithms like linear regression, decision
trees, random forests, neural networks, etc.

6. Model Training:

● Split the data into training and validation sets to train and evaluate the model's
performance.

● Fine-tune hyperparameters to optimize model performance.

● Implement cross-validation techniques to assess model robustness.

7. Model Evaluation:

● Use evaluation metrics specific to your problem (e.g., accuracy, F1-score, RMSE) to
measure how well the model performs.

● Consider confusion matrices, ROC curves, and precision-recall curves for classification
problems.

8. Model Interpretability:

● Ensure that your predictive model is interpretable, especially in business settings where
decision-makers need to understand the model's reasoning.

● Use techniques like feature importance, SHAP values, or LIME to explain model
predictions.

9. Deployment and Monitoring:

● Deploy the model into a production environment for real-world use.

● Continuously monitor the model's performance, retraining it as needed to account for

changing data patterns.

10. Ethical Considerations:

● Be aware of potential biases in the data and models. Address bias and fairness issues to
ensure ethical and responsible use of predictive models.
11. Documentation and Communication:

● Maintain comprehensive documentation of the modeling process, including data

sources, preprocessing steps, and model parameters.

● Communicate the model's findings and insights effectively to non-technical

stakeholders.

Supervised learning

Supervised learning, also known as supervised machine learning, is defined by its use of
labelled datasets to train algorithms that to classify data or predict outcomes accurately.

As input data is fed into the model, it adjusts its weights until the model has been fitted
appropriately. This occurs as part of the cross validation process to ensure that the model
avoids over fitting or under fitting.

Supervised learning helps organizations solve for a variety of real-world problems at scale, such
as classifying spam in a separate folder from your inbox. Some methods used in supervised
learning include neural networks, naïve bayes, linear regression, logistic regression, random
forest, support vector machine (SVM), and more. It offers several significant advantages:

Predictive Power: Supervised ML models can accurately predict outcomes, aiding in

forecasting sales, demand, and customer behavior.

Optimized Decision-Making: It enables data-driven decision-making, optimizing strategies for

marketing, pricing, and resource allocation.

Customer Insights: Supervised ML helps uncover valuable insights into customer preferences,
allowing businesses to tailor products and services.

Risk Assessment: It's instrumental in identifying potential risks and fraud through anomaly
detection, enhancing security and financial management.

Automation and Efficiency: Automation of routine tasks and processes leads to increased
operational efficiency and cost savings.

Personalization: Businesses can deliver highly personalized experiences, enhancing customer

satisfaction and loyalty.

Competitive Advantage: Organizations that harness supervised ML gain a competitive edge by

staying ahead of market trends and competition.
Continuous Improvement: ML models learn and adapt over time, contributing to continuous
improvement and adaptability in dynamic markets.

Resource Optimization: It aids in optimizing resource allocation, from inventory management

to supply chain logistics.

Real-Time Decision Support: Supervised ML provides real-time insights, enabling quicker and
more informed decisions.