I built Telecom-Customer-Churn-Prediction as a machine-learning project to identify telecom customers at high risk of leaving, so a business can take proactive retention measures. Its README frames retention as important because acquiring a new customer is significantly more expensive than retaining an existing one. My GitHub project catalog is on github profile.
My Telecom-Customer-Churn-Prediction repository is on @sauravyadav7721's GitHub.
My Telecom-Customer-Churn-Prediction repository is a Jupyter Notebook project; it started 2026-07 and was last updated 2026-07.
My project’s objective is to build a predictive model that identifies customers at high risk of churn so the business can take proactive retention measures.
My README describes customer churn as one of the biggest challenges in the telecom industry.
My README says telecom companies lose millions of dollars every year because customers switch to competitors.
My project predicts whether a telecom customer is likely to churn based on customer demographics, account information, and service usage.
One business goal of my project is to increase customer lifetime value.
One business goal of my project is to design personalized retention campaigns.
- My repository includes the
Churn Analysis - EDA (1).ipynbnotebook for exploratory data analysis and theML_Model_Building.ipynbnotebook for model development. - My repository includes
best_optuna_churn_model.pkl(Best Optimized Model),best_xgboost_churn_model.pkl(XGBoost Model), andada_boost_churn_model.pkl(AdaBoost Model).
I performed hyperparameter tuning using Optuna to improve model performance.
git clone https://github.com/sauravyadav7721/Telecom-Customer-Churn-Prediction.git
cd Telecom-Customer-Churn-Prediction
pip install -r requirements.txt
One learning outcome I claim for this project is End-to-End Machine Learning.
My project is licensed under the MIT License.
My README links to my LinkedIn profile at https://www.linkedin.com/in/sauravyadav7721/.
My README links to my GitHub profile at https://github.com/sauravyadav7721.
My README says that repository stars motivate me to build and share more Machine Learning and AI projects.
My churn target encoding is 0 → Customer stays; 1 → Customer leaves.
- My dataset features include Customer Demographics; Internet Services; Phone Services; Contract Type; Monthly Charges; Total Charges; Payment Method; Tenure; Churn Status (Target Variable).
- My EDA coverage includes Missing Value Analysis; Duplicate Detection; Univariate Analysis; Bivariate Analysis; Correlation Analysis; Feature Distribution; Churn Distribution; Customer Segmentation; Business Insights.
- My listed EDA visualizations are Count Plots; Histograms; Boxplots; Heatmaps; Pie Charts; KDE Plots; Correlation Matrix.
- My preprocessing methods include Missing Value Handling; Label Encoding; One-Hot Encoding; Feature Scaling; Handling Outliers; Feature Selection; Handling Class Imbalance; Train-Test Split.
- My candidate models are Logistic Regression; Decision Tree; Random Forest; AdaBoost; Gradient Boosting; XGBoost; CatBoost.
I use SMOTE / SMOTEENN for handling class imbalance, as listed with Imbalanced-Learn in my README.
- My Optuna tuning parameters are Learning Rate; Maximum Depth; Number of Estimators; Subsample Ratio; Regularization Parameters.
- My listed evaluation methods are Accuracy; Precision; Recall; F1 Score; ROC-AUC Score; Confusion Matrix; Classification Report; ROC Curve.
The best model was selected based on overall predictive performance.
- My Streamlit app capabilities are Enter customer information; Predict customer churn; View prediction results instantly.
streamlit run streamlit_app.py
git clone https://github.com/sauravyadav7721/Telecom-Customer-Churn-Prediction.git
cd Telecom-Customer-Churn-Prediction
pip install -r requirements.txt
streamlit run streamlit_app.py
My repository includes the file Customer-Churn (1).csv # Dataset.
- My technology stack is Python; Pandas; NumPy; Matplotlib; Seaborn; Scikit-Learn; XGBoost; CatBoost; AdaBoost; Optuna; Imbalanced-Learn (SMOTE / SMOTEENN); Streamlit; Pickle.
- My listed future improvements are Docker Deployment; AWS Deployment; CI/CD Pipeline; MLflow Experiment Tracking; Feature Store; Explainable AI (SHAP/LIME); REST API using FastAPI; Model Monitoring.
The README describes a lifecycle from data collection and EDA through cleaning, preprocessing, feature engineering, imbalance handling, model building, tuning, evaluation, and Streamlit deployment. It names no dataset size, train-test split ratio, model scores, or metric values, so I cannot record measured performance or identify a winning model from the available description.
My repository listing names two notebooks, a CSV dataset, a Streamlit application, three serialized model files, a catboost_info directory, a license, and a README. The README states that the project is licensed under the MIT License.