EthanGabis/Banking_Customer_Churn_Prediction_Dataset

Assignment #1: EDA & Dataset - Bank Customer Churn Analysis

---

Overview

This project presents an exploratory data analysis (EDA) of a bank customer churn dataset. The analysis aims to uncover key factors influencing customer churn and provide actionable insights for customer retention strategies.

Dataset Information

• Source: Bank Customer Churn Dataset
• Size: 10,000 records
• Features: Multiple demographic and behavioral attributes including age, credit score, balance, number of products, and activity status
• Target Variable: Exited (whether a customer churned or not)

Project Structure

The analysis follows a structured approach covering:

1. Data Loading and Initial Exploration

   • Dataset inspection and basic statistics
   • Initial data quality assessment

2. Data Cleaning and Preprocessing

   • Handling missing values
   • Data type conversions
   • Feature scaling and transformation

3. Outlier Detection and Handling

   • IQR (Interquartile Range) method applied to numerical features
   • Analysis of outliers in: CreditScore, Age, Tenure, Balance, EstimatedSalary, and NumOfProducts
   • 

4. Exploratory Data Analysis

   • Descriptive statistics
   • Distribution analysis of key features
   • Correlation analysis

5. Data Visualization

   • Churn rate analysis by age groups
   • Active member status impact on churn
   • Product holdings vs. churn rate
   • Correlation matrix heatmap

Key Findings

Outlier Analysis

• Result: No significant outliers were detected using the IQR method
• Implication: The dataset shows a robust distribution across all numerical features
• Dataset Integrity: All 10,000 records were retained for analysis

Churn Statistics

• Overall Churn Rate: 20.37% of customers in the dataset
• Churn Distribution: Approximately 2,037 customers churned out of 10,000

Critical Churn Factors

1. Age Demographics

• High-Risk Group: Customers aged 40-60 show significantly higher churn rates
• Low-Risk Groups: Both younger customers (under 30) and older customers (above 60) demonstrate greater stability
• Actionable Insight: Targeted retention strategies should focus on the mid-40s to late 50s demographic

2. Customer Activity Status

• Key Finding: Inactive members represent a high-risk group for churn
• Correlation: Weak negative correlation (-0.1561) between active membership and churn
• Recommendation: Engagement strategies are crucial to maintain customer involvement

3. Product Holdings

• Optimal Range: Customers with 1-2 products show the lowest churn rates
• Surprising Pattern: Customers holding 3-4 products exhibit unexpectedly higher churn rates
• Positive Correlation: 0.1854 correlation between number of products and churn
• Investigation Needed: This counterintuitive finding suggests potential issues with:
  • Product utility for multi-product holders
  • Customer service quality for complex portfolios
  • Aggressive cross-selling practices that may backfire

4. Other Notable Correlations

• Age: Moderate positive correlation (0.3124) with churn
• Balance: Weak positive correlation (0.1185) with churn
• Credit Score: Negligible linear relationship with churn
• Tenure: Very weak correlation with churn
• Estimated Salary: Negligible impact on churn

Data Processing Highlights

Feature Scaling

• Successfully applied StandardScaler to numerical features
• Achieved means close to 0 and standard deviations close to 1
• Maintained original values for interpretability in visualizations

Data Quality

• Missing Values: Properly handled (if any were present)
• Data Integrity: All records maintained after outlier analysis
• Feature Engineering: Original product counts preserved alongside scaled versions

Visualizations Included

1. Age Distribution and Churn Analysis: Bar plot showing churn rates across age groups
2. Active Member Status Impact: Comparative analysis of churn between active and inactive members
3. Product Holdings Analysis: Churn rate visualization by number of products held
4. Correlation Matrix Heatmap: Comprehensive view of relationships between all numerical features

Technologies Used

• Python 3
• pandas: Data manipulation and analysis
• numpy: Numerical computations
• matplotlib: Data visualization
• seaborn: Statistical data visualization
• sklearn: Data preprocessing (StandardScaler)

Business Recommendations

Based on the analysis, the following retention strategies are recommended:

1. Age-Targeted Programs: Develop specialized retention programs for customers aged 40-60
2. Engagement Initiatives: Implement proactive engagement campaigns for inactive members
3. Product Portfolio Review:
   • Investigate why multi-product (3-4) customers churn at higher rates
   • Review customer service quality for complex product portfolios
   • Reassess cross-selling strategies
4. Balance-Based Interventions: Consider targeted offers for customers with higher balances who show churn risk

Churn Analysis by Age

Churn Analysis by Active Member

Churn Analysis by Number of Products

Correlation HeatMap

How to Use This Analysis

1. Review the Jupyter notebook for detailed code and step-by-step analysis
2. Watch the video presentation for a quick overview of findings
3. Examine visualizations to understand churn patterns
4. Apply the business recommendations to develop targeted retention strategies

Author

Ethan Gabis
ID: 209926781

Video Presentation

https://youtu.be/3QpimSUF6BI

Acknowledgments

This analysis was completed as part of Assignment #1 for the Data Science course, focusing on building foundational EDA skills and communicating data-driven insights effectively.

---

Last Updated: November 2025