🔄 Telco Customer Churn Prediction | End-to-End ML Pipeline 🚀

Built a production-ready machine learning pipeline using Scikit-learn to predict customer churn. Includes full preprocessing, model tuning with GridSearchCV, evaluation, and pipeline export with joblib.

A complete machine learning pipeline built using Scikit-learn to predict customer churn based on subscription and service usage data from a telecom company. This project demonstrates how to construct a reusable, scalable, and production-ready ML system using real-world practices.

🎯 Objective

The objective of this task is to:

• Predict whether a customer will churn (leave the service) or not.
• Build a clean and modular ML pipeline using Pipeline and ColumnTransformer.
• Perform data preprocessing, model training, hyperparameter tuning, and model export.
• Make the model reusable and ready for deployment using joblib.

🧠 Methodology / Approach

1. Data Loading & Cleaning

• Used the Telco Customer Churn Dataset.
• Cleaned missing values and dropped non-informative columns.
• Converted Churn target to binary labels (Yes → 1, No → 0).

2. Train-Test Split

• Split data into training (80%) and testing (20%) sets using stratification to preserve class balance.

3. Preprocessing with Pipelines

• Used ColumnTransformer to apply:
• StandardScaler to numerical features.
• OneHotEncoder to categorical features.
• Created clean pipelines for preprocessing using Pipeline.

4. Modeling

• Built two machine learning models:
• Logistic Regression
• Random Forest Classifier
• Combined models with preprocessing pipelines for modularity.

5. Hyperparameter Tuning

• Used GridSearchCV to find optimal hyperparameters:
• C for Logistic Regression
• n_estimators, max_depth for Random Forest

6. Model Evaluation

• Evaluated on accuracy, precision, recall, and F1-score using classification_report.
• Compared performance of both models on the test set.

7. Model Export

• Exported the best model pipeline (Random Forest) using joblib for reuse and deployment.

📊 Key Results / Observations

Model	Accuracy (Test Set)	Observations
Logistic Regression	~80%	Simpler, linear model
Random Forest	~83%	Better performance due to non-linearity

• Top churn indicators: month-to-month contract, high monthly charges, low tenure, lack of tech support.
• Random Forest model selected as the final model due to superior performance.

💾 Reusability:

To load and use the saved pipeline:

import joblib

Load saved model

model = joblib.load('telco_churn_pipeline.pkl')

Predict on new data

predictions = model.predict(X_new)

🚀 Future Improvements:

Add feature importance and visual explanations

Deploy the model using Flask or Streamlit

Build a REST API for real-time predictions

Monitor model performance over time (production monitoring)

📚 Tech Stack:

Python 3

pandas, numpy, matplotlib, seaborn

scikit-learn (Pipeline, ColumnTransformer, GridSearchCV, joblib)

Google Colab

📥 Dataset:

Telco Customer Churn Dataset

Source: Kaggle

Link: https://www.kaggle.com/datasets/blastchar/telco-customer-churn