Predictive Analytics for Customer Retention

Welcome to the Predictive Analytics for Customer Retention project! This repository showcases a machine learning solution designed to address customer churn in the telecom sector by predicting high-risk customers and providing actionable insights to minimize churn rates.

Project Objectives

1. Identify Churn Drivers
   Conduct Exploratory Data Analysis (EDA) to uncover key factors influencing customer churn.

2. Build Predictive Models
   Develop and optimize supervised machine learning models to classify customers based on their churn likelihood.

3. Propose Retention Strategies
   Leverage model insights to recommend effective retention strategies for telecom companies.

---

Features

• Exploratory Data Analysis (EDA)

  • Cleaned and preprocessed customer datasets.
  • Identified and visualized key churn indicators.

• Machine Learning Models

  • Implemented Logistic Regression, Decision Trees, and Random Forest Classifier.
  • Applied hyperparameter tuning to enhance performance metrics.

• Evaluation Metrics

  • Evaluated models using confusion matrices, ROC curves, and AUROC scores.

• Business Insights

  • Proposed targeted retention strategies based on churn factors.

---

Tools and Technologies

• Programming Language: Python
• Libraries Used: Pandas, Matplotlib, scikit-learn
• Evaluation Metrics: Confusion Matrices, ROC Curves, AUROC Score

---

Key Results

• Feature Optimization: Reduced irrelevant features by 30%, focusing on impactful churn indicators.
• Improved Recall: Enhanced Random Forest recall from 48% to 65% through hyperparameter tuning.
• Model Performance: Achieved an AUROC score of 0.8514, ensuring reliable churn predictions.