ai ml presentation.pptx ON SUBSCRIPTION BASED INDUSTRY

SUBSCRIBTION BASED INDUSTRY

1. “ Customer Churn Prediction Using Machine Learning ” AI/ML Presentation MBA 2023-2025 Presented by: Dimpy Agarwal (2352110) Prerna Agarwal (2352301) Saransh Aggarwal (2352324) Mahak Goyal (2352141) Prerna Agarwal (2352301)

2. Predicting Customer Churn Using Machine Learning Models: Logistic Regression, KNN, SVM, and Decision Tree Predicting customer churn is a crucial challenge for businesses, particularly in subscription-based industries. By anticipating customer churn and taking proactive steps, companies can prevent losses, retain customers longer, and improve overall performance. This project aims to tackle this challenge using various machine learning techniques to build accurate churn prediction models. Project Title

3. The objectives of this project are as follows T o preprocess the data by handling categorical features and scaling numerical features. To build machine learning models (Logistic Regression, KNN, SVM, Decision Tree) to predict customer churn. To evaluate the performance of these models using accuracy and classification metrics. To use data visualization techniques like scatter plots and heatmaps to understand feature relationships. To make business-driven conclusions based on the results obtained from these models. Objectives

4. Age : Customer's age. Gender : Male/Female. Tenure : Duration of subscription. Usage Frequency : How often the service is used. Support Calls : Number of customer support interactions. Payment Delay : Days of delayed payments. Subscription Type : Basic, Standard, Premium. Contract Length : Monthly, Quarterly, Annual. Total Spend : Total money spent by the customer. Last Interaction : Days since last interaction. Churn : Whether the customer churned (1 = Yes, 0 = No). Dataset Overview

5. Step 1 Data Preprocessing Drop Irrelevant Columns : Removed Customer ID as it doesn’t contribute to churn prediction. Label Encoding : Converted categorical variables (e.g., Gender, Subscription Type) into numerical values for model compatibility. Feature Scaling : Used Standard Scaler to standardize numerical features like Age, Tenure, and Total Spend, ensuring all features are on a similar scale. Handling Missing Values : Imputed missing data to maintain model accuracy. Step 3 Model Building Logistic Regression: A linear model for binary classification. K-Nearest Neighbours (KNN): Classifies data points based on nearest neighbors. Support Vector Machine (SVM): Separates data by finding the best hyperplane . Decision Tree : Splits the data into branches for easy interpretation. Step 2 Data Visualization Scatter Plot : Shows the relationship between tenure and total spend for churned vs. non-churned customers. Heatmap : Displays correlations between different features in the dataset. METHODOLOGY

6. Step 5 - Decision Tree Splitting Data : Divides the dataset into branches based on feature values. Node Decisions : Each internal node represents a decision based on a feature condition. Training : Built using the Gini Index to minimize impurity at each split. Leaf Nodes : Represent the final prediction (Churn or No Churn). Visualization: Generated a decision tree diagram using Scikit- learn’s plot_tree function. METHODOLOGY Step 4 -Model Evaluation Accuracy Score: Evaluated each model based on its accuracy . Classification Report: Assessed precision, recall, and F1-score to measure performance. Visualization: Decision Tree visualized using Scikit- learn's plot tree function.

7. Output

8. Results Logistic RegressLogistic Regression : 83% accuracy K-Nearest Neighbours (KNN) : 91% accuracy Support Vector Machine (SVM) : 94% accuracy (Best performance) Decision Tree : 87% accuracy, interpretable but less accurate accurate

9. Conclusion

10. THANK YOU