I Built an AI That Predicts Who Will Cancel Netflix | Project StreamGuard

📌 ABOUT THE PROJECT Customer churn (subscribers cancelling) is one of the biggest threats to a streaming business, because winning a new customer costs far more than keeping an existing one. StreamGuard is an end-to-end ML pipeline with two goals: 1. Classification: predict whether a subscriber will churn (yes/no). 2. Regression: forecast each customer's Customer Lifetime Value (CLV) in USD. The dataset is synthetic (generated for this project) and modelled on a Netflix-style service, with deliberately messy data: missing values, a $99,999 fee typo, and negative watch hours. ━━━━━━━━━━━━━━━━━━━━ 🛠️ WHAT I BUILT Data cleaning: median and mode imputation, IQR outlier detection and capping Feature engineering: Engagement Efficiency Index (watch hours per dollar) and Engagement Index One-Hot Encoding and Z-Score standardization, fitted only on training data to avoid data leakage Recursive Feature Elimination with Cross-Validation (RFECV) Stratified 5-Fold cross-validation Baseline models: Logistic Regression, KNN, SVM Ensemble models: Random Forest (bagging) and AdaBoost (boosting), tuned with RandomizedSearchCV CLV regression: Linear Regression, Random Forest and Gradient Boosting ━━━━━━━━━━━━━━━━━━━━ 📊 KEY RESULTS Churn prediction (best model: Logistic Regression) Cross-validated ROC-AUC: 0.756 Held-out test set (120 customers): 87% recall, 0.84 ROC-AUC, 77.5% accuracy Surprise: the simple baseline beat the tuned ensembles on this dataset CLV forecasting (best model: Gradient Boosting) Test MAE: about $55 | RMSE: about $76 | R²: 0.88 Note: the test set is small (120 rows), so cross-validation scores are the more reliable estimate. ━━━━━━━━━━━━━━━━━━━━ 📚 TERMS EXPLAINED IN THE VIDEO Churn, CLV, cross-validation, stratification, recall, precision, ROC-AUC, MAE, RMSE, R², bagging vs boosting, data leakage ━━━━━━━━━━━━━━━━━━━━ 🧰 TECH STACK Python, pandas, NumPy, scikit-learn, matplotlib, seaborn, Jupyter Notebook, VS Code ━━━━━━━━━━━━━━━━━━━━ ▶️ RUN IT YOURSELF 1. Clone the repo: https://github.com/Indranil-007/Project_St... 2. Install the dependencies: pip install -r requirements.txt 3. Run the pipeline: python main.py 4. Or open notebooks/streamguard.ipynb and Run All ━━━━━━━━━━━━━━━━━━━━ 🔗 CONNECT LinkedIn: https://www.linkedin.com/in/indranil-garai/ GitHub profile: https://github.com/Indranil-007 If this helped you, like the video and subscribe for more data science and ML projects. Questions or ideas? Leave a comment! #MachineLearning #ChurnPrediction #DataScience #Python #ScikitLearn