📉 Telco Customer Churn Prediction

Analyzing customer behavior and building a machine learning model to predict churn — helping telecom businesses retain customers before it's too late.

---

📌 Project Overview

Customer churn is one of the most critical challenges in the telecom industry. This project digs into real customer data to uncover why customers leave and builds a predictive model to identify at-risk customers early, enabling data-driven retention strategies.

---

🎯 Objectives

• Analyze customer data to understand churn patterns
• Identify key factors driving customer churn
• Build and evaluate ML models to classify churn vs. non-churn
• Deliver actionable business recommendations to reduce churn

---

📂 Dataset

Property	Details
Source	Telco Customer Churn Dataset
Records	~7,000 customers
Key Features	Tenure, monthly charges, contract type, internet service, payment method, and more

---

🧹 Data Preprocessing

• Handled missing values in TotalCharges
• Removed duplicate records
• Corrected data types for numerical columns
• Encoded categorical variables for model compatibility

---

📊 Exploratory Data Analysis

Key patterns discovered in the data:

Insight	Finding
🕐 Early churn risk	~50% of churn occurs within the first year
💸 Pricing pressure	Customers with higher monthly charges churn more
📄 Contract type	Month-to-month contracts show the highest churn rate
🌐 Internet service	Fiber optic users show unexpectedly high churn
🛡️ Support services	Customers without tech support or security services churn more

---

🤖 Models Built

• Logistic Regression — interpretable baseline model
• Random Forest — ensemble model for improved performance

⚙️ Techniques Applied

• Feature scaling
• Class imbalance handling via class_weight='balanced'
• Threshold tuning to optimize recall for the churn class

---

📈 Model Performance

Metric	Result
Accuracy	~79%
Recall (Churn)	Improved via class balancing and threshold tuning

Recall was prioritized for the churn class — it's more costly to miss a churning customer than to flag a loyal one.

---

💡 Key Business Insights

1. First-year experience is critical — the majority of churn happens early in the customer lifecycle
2. Pricing dissatisfaction is a significant churn trigger
3. Absence of value-added services (tech support, online security) weakens retention
4. Flexible/short-term contracts increase churn risk compared to annual or two-year plans
5. Payment friction may be contributing to voluntary churn

---

📋 Business Recommendations

• 🚀 Strengthen onboarding — invest in first-year customer engagement programs
• 💰 Offer targeted discounts — reward high-value or at-risk customers proactively
• 🛡️ Bundle support services — promote tech support and security add-ons
• 📅 Incentivize long-term contracts — discounts or perks for annual commitments
• 💳 Improve payment experience — simplify billing and diversify payment options

---

🛠️ Tech Stack

Library	Purpose
pandas / numpy	Data manipulation and analysis
seaborn / matplotlib	Data visualization
scikit-learn	Model building, evaluation, and preprocessing

---

🚀 Getting Started

# Clone the repository
git clone https://github.com/akshaypal912/telco-churn-prediction.git
cd telco-churn-prediction
 
# Install dependencies
pip install -r requirements.txt
 
# Launch notebooks
jupyter notebook

---

📄 License

This project is open source and available under the MIT License.