Prediction Effect

Prediction effect refers to the impact or influence that a prediction has on outcomes or decisions. In
data analytics, it shows how well a model’s predictions explain or affect the target outcome. In
psychology or society, it means how people’s behavior changes because of a prediction. In short:
Prediction effect = the consequence that predictions themselves create, either on results, decisions,
or behavior.

Deployment of Prediction Model

Deployment of a prediction model means taking a trained machine learning (ML) model and making
it available for real-world use. Steps include: 1. Model Training & Validation – Build and test the
model. 2. Save the Model – Export for reuse. 3. Choose Deployment Environment – Cloud, server,
or edge device. 4. Expose the Model – Create an API or integrate into applications. 5. Monitoring &
Maintenance – Track performance and update as needed. Example: A bank deploys a loan
prediction model via an API to evaluate loan risk automatically.

Ethics and Responsibilities of Predictive Models

Ethical issues: 1. Bias & Fairness 2. Transparency 3. Privacy 4. Accountability 5. Misuse of
Predictions Responsibilities: 1. Ensure Data Quality 2. Fairness & Inclusion 3. Transparency &
Explainability 4. Privacy Protection 5. Continuous Monitoring 6. Human Oversight In short:
Predictive models must be fair, transparent, accountable, and privacy-respecting.

Data Effect
Data effect refers to the influence that data has on outcomes, decisions, or models. Good quality
data improves accuracy and decision-making. Poor or biased data can mislead predictions. In
short: The data effect is the impact that data creates on models, decisions, and society.

Predictive Analysis
Predictive analysis uses historical data, statistics, and machine learning to forecast future
outcomes. Steps: Data Collection → Cleaning → Model Building → Prediction → Evaluation →
Deployment. Applications: Business forecasting, healthcare, finance, and manufacturing.

Types of Predictive Analysis

1. Predictive Modeling 2. Classification Analysis 3. Regression Analysis 4. Time Series Forecasting
5. Clustering & Pattern Recognition 6. Outlier Detection 7. Decision Trees & Rule-Based Prediction
Each type is chosen depending on whether the goal is to predict a number, a category, or detect
patterns.

Types of Analytics
1. Descriptive Analytics – What happened? 2. Diagnostic Analytics – Why did it happen? 3.
Predictive Analytics – What will happen? 4. Prescriptive Analytics – What should we do? 5.
Cognitive Analytics – What can we learn and automate? In short: Descriptive → Diagnostic →
Predictive → Prescriptive.

Predictive Modelling – Decision Trees

Decision Trees are used for classification and regression problems. They split data into branches
based on conditions to make predictions. Advantages: Simple, interpretable, handles
numerical/categorical data. Disadvantages: Overfitting and sensitivity to small data changes.
Example: Loan approval based on income and credit score.

Predictive Analytics Processing Steps

Steps: 1. Problem Definition 2. Data Collection 3. Data Cleaning & Preparation 4. Data Exploration
& Analysis 5. Model Building 6. Model Evaluation 7. Model Deployment 8. Monitoring &
Maintenance In short: Problem → Data → Clean → Analyze → Build → Evaluate → Deploy →
Monitor.

Setting Up the Problem: Processing Steps

1. Setting Up the Problem – Define objective, context, target variable, and constraints. 2. Predictive
Analytics Processing Steps – Data collection, cleaning, exploration, model building, evaluation,
deployment, and monitoring. In short: Problem setup defines what to predict, then the predictive
process executes how to do it.

Business Understanding: Objectives

Business understanding defines business goals before modeling. Objectives: 1. Define business
problem. 2. Identify goals & success criteria. 3. Understand processes & constraints. 4. Identify
stakeholders. 5. Convert to analytical problem. 6. Determine data needs. 7. Define expected
outcomes. Example: Telecom company predicting customer churn to reduce loss.

Data for Predictive Modeling

Types of Data: 1. Historical 2. Real-time 3. Structured 4. Unstructured 5. Semi-structured Sources:
Internal (CRM, databases) & External (web, IoT, government data). Preparation includes cleaning,
transformation, and feature selection. In short: Quality data = Accurate predictions.

Measures of Success for Predictive Models

Classification Models: - Accuracy, Precision, Recall, F1 Score, ROC-AUC. Regression Models: -
MAE, MSE, RMSE, R². Business Measures: - Cost saving, efficiency, fairness. In short: Evaluate
model using performance metrics and business impact.

Setting Up the Problem: Columns as Measures in Predictive

Modeling
Columns represent features and target variables. Independent Variables – Input features (Age,
Salary, Credit Score). Dependent Variable – Target output (Default Yes/No). Derived Measures –
Calculated columns (Total Purchase = Qty × Price). In short: Setting up the problem involves
selecting correct input and target columns for accurate prediction.