Customer Segmentation & Churn Prediction

📌 Project Overview

This project aims to segment e-commerce customers based on their behavior and predict their churn probability using clustering and classification techniques.

📊 Key Statistical Concepts Used

• Descriptive Statistics & Correlation Analysis
• Dimensionality Reduction: PCA, t-SNE
• Clustering: K-Means (faiss), Gaussian Mixture Model (GMM)
• Predictive Modeling: Logistic Regression, Random Forest
• Model Evaluation: Confusion Matrix

🔗 Dataset

• Source: E-Commerce Customer Churn Analysis Dataset

📌 Steps Followed

1️⃣ Data Preprocessing

• Handled missing values using Iterative Imputer.
• Standardized numerical features using Standard Scaler.

2️⃣ Exploratory Data Analysis (EDA)

• Performed univariate and bivariate analysis.
• Visualized missing values using missingno.

3️⃣ Customer Segmentation

• Applied Faiss K-Means for efficient clustering.
• Used Gaussian Mixture Model (GMM) for probabilistic clustering.
• Evaluated clusters using Silhouette Score and Calinski-Harabasz Score.

4️⃣ Churn Prediction Model

• Trained Logistic Regression and Random Forest classifiers.

5️⃣ Model Evaluation

• Assessed performance with confusion matrix

🚀 Results

Model	Accuracy
Logistic Regression	84%
Random Forest	96%

🔧 Improvements to be made

• More data visualization
• Implement XGBoost for better performance.
• Hyper parametertune Random Forest for optimal parameters.

📂 Project Structure

• Customer_Clustering_and_Churn_Prediction.py: code file
• Churn_predictions_gmm.csv: predictions