Customer Churn Prediction System 📊

A Machine Learning based web application that predicts customer churn probability using customer subscription and behavioral data.

The project focuses on identifying high-risk customers early using classification techniques and recall optimization.

---

🚀 Features

• Predicts customer churn probability
• Logistic Regression based ML model
• Recall focused threshold optimization
• Interactive Streamlit dashboard
• Customer risk classification
• Churn risk visualization
• Retention recommendation system

---

🧠 Machine Learning Workflow

1. Data Cleaning
2. Exploratory Data Analysis
3. Feature Encoding
4. Feature Scaling
5. Logistic Regression Training
6. Threshold Optimization
7. Model Deployment using Streamlit

---

📊 Dataset

IBM Telco Customer Churn Dataset

Dataset contains:

• 7043 customer records
• Demographic information
• Subscription details
• Billing information
• Customer churn labels

---

📈 Model Performance

Metric	Score
Accuracy	71%
Recall	81%

Recall optimization is applied to reduce false negatives and improve detection of customers likely to churn.

---

🛠 Tech Stack

• Python
• Pandas
• NumPy
• Scikit-learn
• Streamlit
• Plotly
• Joblib

---

📂 Project Structure

Customer-Churn-Prediction-System

├── data
│   └── Telco-Customer-Churn.csv
│
├── models
│   ├── churn_model.pkl
│   ├── scaler.pkl
│   └── features.pkl
│
├── app.py
├── train_model.py
├── preprocessing.py
├── requirements.txt
└── README.md

---

▶️ Run Locally

Clone repository

git clone <repo-link>

Install dependencies

pip install -r requirements.txt

Train model

python train_model.py

Start Streamlit App

streamlit run app.py

---

Future Improvements

• Add advanced ML models
• Add SHAP explainability
• Deploy using Streamlit Cloud