customer-churn-prediction-and-risk-segmentation

📉 Customer Churn Prediction & Risk Segmentation

Week 2 Project | AI & Data Analyst Internship

An end-to-end machine learning pipeline that predicts customer churn for a telecom business, segments customers into risk tiers, and delivers business-ready insights through interactive visualizations.

---

📌 Problem Statement

Subscription-based businesses lose significant revenue every year due to customer churn — customers who cancel or stop using a service. Predicting which customers are likely to churn before they actually leave allows the business to act proactively: offer discounts, improve support, or tailor communication.

This project simulates a real analyst task at a product or growth team, with a focus on interpretability — so that a business stakeholder can understand the findings, not just the code.

---

📁 Repository Structure

customer-churn-prediction-and-risk-segmentation/
│
├── Analysis.ipynb            # Main Colab Notebook (all 6 tasks)
├── Model_Comparison.png      # Side-by-side model performance comparison chart
├── requirements.txt          # All required Python libraries
├── summary.pdf               # Final insights & business recommendations report
├── Dataset/                  # Raw telecom customer churn CSV dataset
└── Charts/                   # All generated visualization charts

---

✅ Tasks Completed

Task 1 — Data Loading & Exploratory Analysis

• Loaded the telecom dataset using Pandas and displayed shape, column types, and first 10 rows
• Identified the target column (Churn — Yes/No) and checked for class imbalance
• Found and handled missing/null values (drop vs. impute decision made)
• Computed summary statistics for numerical columns (mean, median, std)
• Plotted a correlation heatmap of numerical features using Seaborn

Task 2 — Data Preprocessing & Feature Engineering

• Converted TotalCharges to numeric with careful coercion error handling
• Encoded all categorical columns using Label Encoding / One-Hot Encoding (with justification)
• Created 2 new engineered features:
  • ChargesPerMonth = TotalCharges / tenure
  • SeniorWithNoSupport = (SeniorCitizen == 1) & (TechSupport == "No")
• Scaled numerical features using StandardScaler
• Split data into 80% train / 20% test using stratified split to preserve class ratio

Task 3 — Model Training & Comparison

• Trained 3 classification models:
  • Model 1: Logistic Regression
  • Model 2: Random Forest Classifier
  • Model 3: XGBoost / Gradient Boosting Classifier
• Evaluated all models with: Accuracy, Precision, Recall, F1-Score, ROC-AUC
• Plotted Confusion Matrix for each model side-by-side
• Plotted ROC Curve for all 3 models on one chart for comparison
• Performed hyperparameter tuning on the best model using GridSearchCV / RandomizedSearchCV (Advanced)

Task 4 — Customer Risk Segmentation (Advanced)

• Used the best model's predicted churn probability (not just 0/1 labels)
• Segmented all customers into 3 risk tiers:
  • 🔴 High Risk → probability ≥ 0.70
  • 🟡 Medium Risk → probability between 0.40 and 0.69
  • 🟢 Low Risk → probability < 0.40
• Counted customers per tier and computed per-tier averages:
  • Average MonthlyCharges, average tenure, and Contract type distribution
• Created a grouped bar chart comparing the 3 tiers across key features

Task 5 — Visualizations (4 Charts)

• Chart 1: Feature importance bar chart (top 10 features) from Random Forest / XGBoost
• Chart 2: Churn rate by Contract Type (Month-to-Month vs. One Year vs. Two Year)
• Chart 3: Distribution of tenure — churned vs. not-churned customers (KDE / overlapping histogram)
• Chart 4: Risk tier donut/pie chart showing customer count per tier
• Bonus Chart: Interactive scatter plot using Plotly — MonthlyCharges vs. Tenure, colored by churn label

All charts are saved in the Charts/ folder. Model comparison chart: Model_Comparison.png

Task 6 — Insights & Business Recommendations

A structured analysis written inside the notebook covering:

• Which model performed best and why it was selected
• Top 3 factors driving customer churn
• What High Risk customers have in common (contract type, charges, tenure)
• 2 specific, actionable business recommendations
• Model limitations and future improvement suggestions

Full written summary available in summary.pdf.

---

🛠️ Tools & Libraries Used

Tool	Purpose
Python 3.x	Main programming language
Google Colab	Development environment
Pandas	Data loading and manipulation
NumPy	Numerical operations
Scikit-learn	Models, metrics, preprocessing
XGBoost	Gradient boosting model
Matplotlib / Seaborn	Static charts and heatmaps
Plotly	Interactive visualizations

Install all dependencies at once:

pip install -r requirements.txt

---

🚀 How to Run This Project

1. Open in Google Colab

Click below to open the notebook directly — no setup needed:

Or manually:

1. Go to colab.research.google.com
2. Click File → Open Notebook → GitHub
3. Paste: https://github.com/smohan-21/customer-churn-prediction-and-risk-segmentation
4. Select Analysis.ipynb

2. Install required libraries (run this cell first in Colab)

!pip install -r requirements.txt

3. Upload or mount the dataset

from google.colab import drive
drive.mount('/content/drive')

Or upload the CSV manually from the Dataset/ folder when prompted.

Run all cells from top to bottom to reproduce the full analysis.

---

📊 Key Outputs

• Cleaned & feature-engineered dataset with churn probability scores
• 3 trained and evaluated ML models with full metrics comparison
• Customer risk segmentation table (High / Medium / Low)
• 4+ visualizations saved in the Charts/ folder
• Model comparison chart: Model_Comparison.png
• Business insights report: summary.pdf

---

👤 Author

Mohan S AI & Data Analyst Intern — Week 2 Project GitHub: @smohan-21

---

📄 License

This project is built for internship and educational purposes.