This project focuses on analyzing customer behavior and building a machine learning model to predict customer churn. The goal is to help businesses identify at-risk customers early and take proactive retention actions.
- Perform Exploratory Data Analysis (EDA) to uncover patterns
- Build predictive models to identify churn customers
- Compare multiple ML algorithms
- Generate actionable business insights
- Age, Gender
- Annual Income
- Spending Score
- Purchase History
- Membership Duration
- Feedback Score
- Churn Indicator (Target Variable)
- Customers with low feedback scores are more likely to churn
- Short membership duration strongly correlates with churn
- Engagement (spending/purchases) impacts retention more than income
- Behavioral factors > demographic factors
- Logistic Regression
- Random Forest
- XGBoost (Final Model)
- XGBoost showed improved churn detection
- Still faced low recall, missing some churn customers
- Highlights need for recall-focused optimization
- Model performs well for non-churn prediction
- Needs improvement in detecting actual churn customers
- Missing churn = potential revenue loss
- Improve customer experience & feedback systems
- Target low-engagement users
- Introduce loyalty & retention programs
- Handle class imbalance (SMOTE / weights)
- Optimize threshold (focus on recall)
- Further model tuning
- Python (Pandas, NumPy)
- Scikit-learn
- XGBoost
- Matplotlib / Seaborn
- Data Cleaning & Preprocessing
- Exploratory Data Analysis
- Feature Engineering
- Model Building
- Evaluation & Optimization
- Insights & Recommendations
- Deploy model with real-time API
- Advanced ensemble models (Stacking)
- Customer segmentation (Clustering)
This project demonstrates how machine learning can be used to predict customer churn and drive business decisions, while highlighting the importance of optimizing models for real-world impact.