Customer Churn Analysis: A Comprehensive Statistical Approach

Overview

This repository contains a complete analysis of the "Churn Modelling" dataset, focusing on understanding customer churn patterns for a bank. The project employs various statistical methods and data visualization techniques to uncover key insights about customer behavior and factors influencing churn.

Table of Contents

1. Introduction
2. Dataset Overview
3. Data Processing
4. Analysis
5. Visualizations
6. Conclusion
7. Usage
8. Requirements

Introduction

Customer churn is a critical metric for businesses, especially in the banking sector. This project aims to analyze the factors that influence whether a customer will close their account with a bank. We utilize statistical methods, data validation, handling missing data, outlier detection, encoding, and principal component analysis (PCA) to gain meaningful insights.

Dataset Overview

The dataset consists of 500 entries with the following features:

• CreditScore: Numerical score representing the customer's credit rating.
• Geography: Country where the customer is located (e.g., France, Spain).
• Gender: Customer’s gender (Male or Female).
• Age: Customer’s age.
• Tenure: Number of years the customer has been with the bank.
• Balance: Balance in the customer's bank account.
• EstimatedSalary: Customer’s estimated annual salary.
• Exited: Binary target variable indicating whether the customer has left the bank (1=exited, 0=remained).

Dataset Source

Data Processing

Steps Undertaken:

1. Loading the Data

   • Using readr library to load the dataset.

2. Data Validation

   • Summary statistics using summary() function.
   • Data profiling using DataExplorer package.

3. Handling Missing Data

   • Identification and imputation of missing values.

4. Encoding Categorical Variables

   • Conversion of categorical variables into numerical representations using factor().

5. Outlier Detection and Removal

   • Identification and removal of outliers using IQR method.

Analysis

Key Findings:

• Credit Score: Customers who did not exit tend to have better credit scores than those who exited.
• Age: Non-exited customers are generally younger, with most common age around 35.
• Tenure: Long-term customers show higher retention rates.
• Balance: Higher balance customers are more likely to leave.
• Estimated Salary: Wide variability observed among both groups.

Statistical Measures:

• Central tendency and dispersion measures were calculated for both exited and non-exited groups.

Visualizations

Various visualizations were created to understand the data better:

1. Box Plots:

   • Comparison of CreditScore, Balance, and EstimatedSalary between exited and non-exited customers.

2. Histograms:

   • Age and Tenure distributions against the Exited variable.

3. Bar Plots:

   • Gender and Geography distributions against the Exited variable.

Conclusion

The analysis revealed several actionable insights:

• Middle-aged and German customers have a higher likelihood of exiting.
• Customers with higher balances and shorter tenures are more likely to leave.
• Targeted interventions for specific customer segments could improve retention efforts.

Usage

To run the code locally, follow these steps:

1. Clone this repository:

   git clone https://github.com/yourusername/customer-churn-analysis.git
   

2. Install the required packages:

   install.packages(c("readr", "DataExplorer", "ggplot2", "reshape2"))
   

3. Run the R script:

   source("path_to_your_script.R")
   

Requirements

R version 4.4.3 or later
Required Libraries: readr, DataExplorer, ggplot2, reshape2