churn.ai — Machine Learning Customer Churn Predictor

An advanced, state-of-the-art customer churn prediction dashboard powered by multiple machine learning models. Built to evaluate customer data in real time, analyze risk factors, and provide tailored proactive outreach recommendations.

🔗 Live Production Demo on Render

---

🌟 Key Features

• Advanced Professional Dashboard: Fully styled with an elegant, modern warm-sand color palette (#FAF9F6), high-end typography (Playfair Display & Plus Jakarta Sans), and custom organic fluid wave gradients.
• Responsive Layout Alignment: An interactive tab-based navigation system (Personal Profile, Phone Services, Internet Services, Billing & Payments) that guarantees flawless grid alignment and prevents uneven elements.
• Aggregate Risk Scoring: Displays a live, animated circular SVG risk ring that aggregates predictions across all models to calculate a unified customer churn risk percentage.
• Proactive Retention Strategies: Automatically recommends tailored marketing or customer success steps based on the calculated risk tier (Low, Moderate, High Risk).
• Multi-Model Pipeline Comparison: Seamlessly computes and compares predictions and confidence scores for four top-tier classification models simultaneously:
  • ❖ Logistic Regression
  • ✦ Decision Tree
  • ✹ Random Forest
  • 💠 Support Vector Machine (SVM)

---

🛠️ Architecture & Technical Stack

graph TD
    A[Interactive Frontend - HTML5/CSS3/JS] -->|POST Payload| B[Flask Backend API - app.py]
    B --> C{ML Model Loader}
    C -->|Logistic Regression| D[log_reg.pkl]
    C -->|Decision Tree| E[decision_tree.pkl]
    C -->|Random Forest| F[random_forest.pkl]
    C -->|SVM| G[svm.pkl]
    B --> H[Standard Scaler & Encoders]
    B -->|Response JSON| A

Loading

Frontend

• Core: HTML5, Vanilla JavaScript (ES6+), and CSS3.
• Styling & Aesthetics: High-contrast modern card views, dynamic hover highlights with premium gold accents (#c69c3a), responsive scrollable pill bars, and interactive selected card animations.

Backend

• Web Server: Python (Flask) with Gunicorn WSGI for robust production deployment.
• CORS: Configured with flask-cors for secure cross-origin queries.

Machine Learning

• Framework: scikit-learn for classification algorithms.
• Data Processing: Pre-trained standard scalers and categorical label encoders persisted via joblib.

---

📂 Project Directory Structure

├── models/
│   ├── log_reg.pkl           # Trained Logistic Regression model
│   ├── decision_tree.pkl     # Trained Decision Tree classifier
│   ├── random_forest.pkl     # Trained Random Forest classifier
│   ├── svm.pkl               # Trained Support Vector Machine model
│   ├── scaler.pkl            # StandardScaler instance for input normalization
│   └── encoders.pkl          # Persisted label encoders for categorical values
├── static/
│   └── index.html            # Redesigned premium frontend dashboard
├── app.py                    # Flask server & model inference routes
├── requirements.txt          # Python dependencies
├── Procfile                  # Startup configuration for cloud platforms
└── README.md                 # Project documentation

---

💻 Local Setup & Development

Prerequisites

Make sure you have Python 3.8+ installed on your machine.

1. Clone the Repository

git clone https://github.com/dinesh9997/Customer-Churn-Prediction.git
cd Customer-Churn-Prediction

2. Create and Activate Virtual Environment

# On Windows
python -m venv venv
venv\Scripts\activate

# On macOS/Linux
python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Start the Application

python app.py

5. View in Browser

Open your browser and navigate to: 👉 http://localhost:5000/

---

🚀 Production Deployment on Render

This project is configured for 1-click deployments on Render.

Render Settings

• Runtime: Python 3
• Build Command: pip install -r requirements.txt
• Start Command: gunicorn app:app
• Instance Type: Free

---

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.

👥 Authors

• Dinesh — Initial Work & Model Engineering (GitHub: @dinesh9997)