Customer Churn Analysis and Prediction

Customer Churn Analysis and Prediction

Project Description: Performed an in-depth analysis of bank customer churn data to identify key drivers of churn and build a predictive model to forecast customer attrition, enabling proactive retention strategies.

Technologies Used: SQL , Python (Pandas, NumPy, Matplotlib, Seaborn, Plotly, SciPy, Statsmodels, Scikit-learn), MS Excel, Power BI.

Achievements:

·        Conducted comprehensive exploratory data analysis (EDA) to understand customer demographics and their relationship with churn, identifying an overall churn rate of 20.37%.

·        Utilized SQL for initial data aggregation and extraction of critical insights, revealing higher churn rates for female customers (25.07%) compared to male customers (16.46%), and significantly higher churn in Germany (32.44%) compared to Spain (16.67%) and France (16.15%).

·        Performed statistical hypothesis testing (Chi-Square and t-tests) to validate significant associations between churn and various customer attributes, including country, gender, age group, active member status, credit score band, product diversity, and customers with "zero balance & high salary".

·        Identified that customers with shorter tenure are more prone to churn, with churned customers having an average tenure of 4 years compared to retained customers with 5 years. The 46-60 (Senior Adult) age group also showed a higher propensity for churn.

·  Developed and implemented a predictive model using a Voting Classifier, achieving a mean cross-validation accuracy of 89.72% and a test accuracy of 85%, demonstrating strong capability in forecasting customer attrition.

https://drive.google.com/drive/u/0/folders/1Sh6ctjETm8aIE9SpZWB_C9BtAmsfv7FL