Exploratory Data Analysis (EDA): Bank Customer Churn Objective The goal of this analysis is to understand the key patterns and drivers behind customer churn in a banking context. Churn — when a customer leaves the bank — can be extremely costly, so uncovering early warning signs is crucial. This EDA serves as the foundation before building any predictive machine learning models. Key Focus Areas Framing the Business Problem: Reducing churn = retaining valuable customers = higher lifetime value. Data Distributions: Analysis of numerical variables (e.g., balance, credit score, age, tenure) to identify skewness, outliers, and transformation needs. Churn Trends by Geography and Demographics: Deep dive into country-level churn (Germany vs France vs Spain), and patterns by gender, age, and tenure. Feature Relationships: Scatter plots, correlation matrix, and churn ratios to explore how variables interact and contribute to churn risk. Handling Imbalanced Data: Churn is typically a minority class — analysis accounts for the skew and explores preprocessing ideas for modeling. Tools Used R & R Markdown tidyverse, ggplot2, DataExplorer, caret Interactive visualizations and summary statistics Key Insights Tenure = 1 year is the most volatile group with the highest churn rate. German customers show significantly higher churn than other countries. Balance and credit score distributions reveal potential high-risk segments. Certain variables are uniformly distributed, offering low predictive power. Class imbalance confirmed — churners are a minority class (~20%), requiring resampling or class-weighted models in future steps. Next Steps Data cleaning and preprocessing (e.g., scaling, encoding) Feature engineering Classification modeling using logistic regression, tree-based methods, and ensemble models This analysis is the first step in building a robust churn prediction model. Proper EDA ensures smarter feature selection, better model accuracy, and ultimately smarter business decisions.