tidal-customer

byC R Renuka

Act as a senior Machine Learning Engineer and Full Stack Developer. Build a complete "Customer Churn Management System" using Machine Learning with the following requirements: 1. OBJECTIVE: Design a system that predicts whether a customer will churn (leave a service) and provides actionable insights to reduce churn. 2. DATASET: - Use a sample telecom/bank/SaaS churn dataset (like Telco Customer Churn dataset). - Include features such as: - Customer demographics (age, gender, location) - Account info (tenure, subscription type, contract) - Usage behavior (monthly charges, usage frequency) - Support interactions (complaints, tickets) - Target variable: Churn (Yes/No) 3. MACHINE LEARNING PIPELINE: - Perform data preprocessing: - Handle missing values - Encode categorical variables - Feature scaling - Perform Exploratory Data Analysis (EDA) with visualizations - Split dataset into train/test - Train multiple models: - Logistic Regression - Random Forest - XGBoost - Compare models using: - Accuracy - Precision, Recall, F1-score - ROC-AUC - Select best model and tune hyperparameters - Save trained model using pickle/joblib 4. CHURN PREDICTION FEATURES: - Predict churn probability for new users - Display risk score (Low / Medium / High) - Show feature importance (why customer may churn) 5. ADVANCED FEATURES (VERY IMPORTANT): - Add explainability using SHAP or LIME - Provide personalized retention strategies: Example: - High charges → Offer discount - Low usage → Send engagement offers - Add alert system for high-risk customers 6. BACKEND: - Use Python (Flask or FastAPI) - Create APIs: - /predict (input customer data → output churn prediction) - /insights (return churn analysis) 7. FRONTEND: - Build using HTML, CSS, JavaScript - Dashboard should include: - Churn prediction form - Charts (churn rate, feature importance) - Customer risk segmentation 8. VISUALIZATION: - Use matplotlib / seaborn / plotly - Show: - Churn distribution - Correlation heatmap - Feature importance graph 9. DEPLOYMENT: - Provide steps to deploy: - Local deployment - Cloud (Render / Railway / AWS) 10. BONUS (MAKE IT STAND OUT): - Add login system (admin dashboard) - Add real-time prediction - Integrate email alerts for churn-risk customers - Use free APIs if needed 11. OUTPUT REQUIRED: - Clean modular code (separate files) - Folder structure - README.md - Sample dataset - Screenshots of dashboard Make the project industry-level, scalable, and resume-worthy (FAANG-level quality).