Ecommerce-churn-prediction

AI-powered e-commerce customer churn prediction web app built with Python, Random Forest and Streamlit

Overview

This project predicts whether a customer will churn or not using machine learning. It analyzes customer behavior like app usage, order patterns, and engagement to identify high-risk customers.

Problem Statement

Customer churn is a major issue for ecommerce companies. This project helps identify customers who are likely to leave so that businesses can take preventive actions.

Dataset

• Ecommerce customer dataset (https://www.kaggle.com/datasets/ankitverma2010/ecommerce-customer-churn-analysis-and-prediction)
• Sheet used: E Comm
• Features include:
  • Tenure
  • Cashback Amount
  • Warehouse distance
  • Order behavior
  • Preferred login device
  • Payment mode

Tools and Technologies

• Python
• Pandas and NumPy
• Matplotlib and Seaborn
• Scikit-learn

Project Workflow

1. Data Loading
2. Data Cleaning (handling missing values using median)
3. Encoding categorical variables
4. Exploratory Data Analysis
5. Model Building using Random Forest
6. Model Evaluation
7. Feature Importance Analysis

Model Used

• Random Forest Classifier

Model Evaluation

• Model Accuracy: 98%+
• Confusion Matrix used for performance evaluation
• Classification Report generated

Key Insights

• Customer tenure and engagement significantly impact churn
• Purchase behavior patterns help identify high-risk customers
• Feature importance highlights key drivers of churn

Churn Prediction System

• Built a function to take customer inputs and predict churn
• Helps classify customers into risk categories

Model Saving

• Model saved as churn_model.pkl using pickle
• Can be reused without retraining

How to Run

1. Upload the dataset (Excel file)
2. Run the notebook or Python file
3. Use prediction function to test new data

Screenshots

Author

ANIKA ATTRI