Customer Churn Prediction

Predict whether a customer will churn (leave) a telecom company using logistic regression.
This project includes data cleaning, exploratory data analysis (EDA), feature engineering, model building, and actionable business insights.

---

📂 Dataset

• Source: Kaggle Telco Customer Churn
• Number of records: 7,043 customers
• Features:
  • Customer demographic info (gender, senior citizen, partner, dependents)
  • Service details (Internet service, phone service, streaming)
  • Billing info (monthly charges, total charges, payment method, contract type)
• Target: Churn (Yes/No)

---

🛠 Key Steps

1. Data Loading & Cleaning

   • Handle missing and inconsistent values
   • Convert TotalCharges to numeric
   • Clean categorical features

2. Exploratory Data Analysis (EDA)

   • Visualized churn distribution
   • Studied relationships between churn and key features:
     • Tenure
     • Contract type
     • Monthly charges
     • Services subscribed

3. Feature Encoding & Engineering

   • Converted categorical features to numeric (One-Hot Encoding & 0/1 mapping)
   • Encoded target variable (ChurnFlag)

4. Model Building

   • Logistic Regression
   • Trained on 80% of data, tested on 20%
   • Evaluated using:
     • Accuracy (~85%)
     • Confusion Matrix
     • ROC-AUC

5. Insights & Recommendations

   • Month-to-month contract customers are most likely to churn
   • High monthly charges slightly increase churn probability
   • Long-tenure customers are less likely to churn
   • Actionable strategies:
     • Encourage month-to-month customers to switch to longer contracts
     • Offer retention incentives for high-paying customers
     • Promote value-added services (Tech Support, Online Security)
     • Onboard new customers proactively

---

📊 Visualizations

• Churn Distribution
• Tenure vs Churn
• Monthly Charges vs Churn
• Contract Type vs Churn
• Feature Importance (Logistic Regression Coefficients)

Tip: Save plots in outputs/ folder and reference in README with