📞 Telecom Customer Churn Prediction

End-to-End Machine Learning Project with Streamlit Deployment

---

📌 Overview

Customer churn is one of the biggest challenges faced by telecom companies. Retaining existing customers is significantly more cost-effective than acquiring new ones.

This project predicts whether a telecom customer is likely to Stay or Churn using Machine Learning.

The project covers the complete Machine Learning lifecycle including:

• Data Cleaning
• Exploratory Data Analysis
• Feature Engineering
• Model Training
• Model Evaluation
• Streamlit Deployment
• Business Recommendation System
• Prediction Report Generation

---

📂 Dataset

Dataset: IBM Telco Customer Churn Dataset

The dataset contains customer information such as:

• Gender
• Senior Citizen
• Internet Service
• Contract Type
• Monthly Charges
• Total Charges
• Payment Method
• Online Security
• Streaming Services

Target Variable:

Churn

---

🚀 Features

✔ Data Cleaning

✔ Exploratory Data Analysis

✔ Feature Engineering

✔ Missing Value Handling

✔ One-Hot Encoding

✔ Feature Scaling

✔ Multiple Machine Learning Models

✔ Hyperparameter Tuning

✔ Customer Churn Prediction

✔ Prediction Probability

✔ Business Recommendations

✔ Feature Importance Visualization

✔ Download Prediction Report

✔ Interactive Streamlit Dashboard

---

🛠 Tech Stack

Category	Technology
Language	Python
Data Analysis	Pandas, NumPy
Visualization	Matplotlib, Seaborn
Machine Learning	Scikit-Learn
Model Saving	Joblib
Web Framework	Streamlit
IDE	VS Code

---

📁 Project Structure

Telecom-Customer-Churn-Prediction/

│
├── models/
│ ├── best_model.pkl
│ ├── scaler.pkl
│ ├── model_columns.pkl
│ └── target_encoder.pkl
│
├── notebooks/
│ ├── EDA.ipynb
│ ├── Preprocessing.ipynb
│ ├── Model_Training.ipynb
│ └── Model_Evaluation.ipynb
│
├── outputs/
│ ├── churn_distribution.png
│ ├── feature_importance.png
│ ├── confusion_matrix.png
│ ├── ROC_curve.png
│ └── ...
│
├── src/
│ ├── load_data.py
│ ├── preprocess.py
│ ├── train_model.py
│ ├── evaluate.py
│ └── predict.py
│
├── streamlit_app/
│ └── app.py
│
├── reports/
│
├── requirements.txt
│
└── README.md

---

⚙ Machine Learning Workflow

Dataset

↓

Data Cleaning

↓

EDA

↓

Preprocessing

↓

Encoding

↓

Scaling

↓

Train/Test Split

↓

Model Training

↓

Model Evaluation

↓

Model Selection

↓

Save Model

↓

Streamlit Deployment

---

📊 Model Performance

Metric	Score
Accuracy	80%+
Precision	High
Recall	High
F1 Score	High

---

💻 Streamlit Dashboard

The application allows users to:

• Enter customer details
• Predict churn probability
• View Stay vs Churn percentages
• Get business recommendations
• View important features
• Download prediction report

---

📸 Screenshots

🏠 Home Page

---

📝 Customer Input

---

📊 Prediction Result

---

💼 Business Recommendation

---

📈 Feature Importance

▶ Installation

Clone the repository

git clone https://github.com/HarshitKumarModi/Telecom-Customer-Churn-Prediction.git

Go inside project

cd Telecom-Customer-Churn-Prediction

Install dependencies

pip install -r requirements.txt

Run Streamlit

streamlit run streamlit_app/app.py

---

📈 Future Improvements

• Deep Learning Models
• XGBoost Integration
• Explainable AI (SHAP)
• Docker Deployment
• Cloud Deployment
• Customer Lifetime Value Prediction

---

👨‍💻 Author

Harshit Kumar Modi

B.Tech Computer Science Engineering

VIT Bhopal University

GitHub: https://github.com/HarshitKumarModi

LinkedIn: https://linkedin.com/in/harshitkumarmodi

---

⭐ If you like this project, don't forget to Star the repository.