📊 Telco Customer Churn Analysis & Prediction

An end-to-end churn project on the Telco Customer Churn dataset: a Power BI dashboard for visual analysis and a Random Forest model in Python to predict which customers are likely to leave, plus business-focused EDA on tenure, contracts, and Lifetime Value (LTV).

---

🎯 What's Inside

File	Description
Customer_Churn-prediction-analysis.ipynb	Full ML notebook — preprocessing, model, evaluation, EDA
Customer-Churn_Telco_Clean.csv	Cleaned dataset used as the model input
Churn_Analysis.pbix	Interactive Power BI dashboard

---

🔬 Modeling Pipeline (notebook)

1. Load the cleaned dataset and drop non-predictive columns (customerID, index).
2. Encode 19 categorical features with LabelEncoder; map the target Churn (No → 0, Yes → 1).
3. Split into features/target with an 80/20 train-test split (random_state=42).
4. Train a RandomForestClassifier (n_estimators=100, random_state=42).
5. Evaluate with a confusion matrix and a full classification report.
6. Rank drivers using the model's feature-importance scores.
7. Engineer a Lifetime Value feature: LTV = MonthlyCharges × tenure.

---

📈 Results

Overall accuracy: 80% (on 1,409 test records)

Metric	Class 0 (Retained)	Class 1 (Churned)
Precision	0.83	0.67
Recall	0.92	0.46
F1-score	0.87	0.55

---

🔑 Key Insights (from the notebook)

• Tenure: churned customers stay ~18 months on average vs ~38 months for retained customers.
• Contract type: month-to-month customers churn far more than one- or two-year contract holders.
• Internet service: fiber-optic users show the highest churn rate; customers with no internet service churn least.
• Lifetime Value: average LTV is $18,266 for churned vs $27,810 for retained customers — churn directly erodes revenue.
• LTV drivers: Contract type and Internet Service influence LTV the most; gender has negligible impact.

---

🛠️ Tools & Libraries

Python · pandas · NumPy · scikit-learn (RandomForestClassifier, LabelEncoder, train_test_split, classification_report, confusion_matrix) · matplotlib · seaborn · joblib · Jupyter Notebook · Power BI (dashboard)

---

🚀 Run It

pip install pandas numpy scikit-learn matplotlib seaborn joblib jupyter
jupyter notebook Customer_Churn-prediction-analysis.ipynb

Open Churn_Analysis.pbix in Power BI Desktop to explore the interactive dashboard.

---

📄 License

MIT