Customer Churn Analysis & Prediction

Project Overview

This project focuses on analyzing customer churn behavior and building machine learning models to predict whether a customer will leave (churn) or stay.

The goal is to identify key factors influencing churn and provide actionable insights that can help businesses improve customer retention.

---

Objectives

• Perform data cleaning and preprocessing
• Conduct exploratory data analysis (EDA)
• Build and evaluate machine learning models
• Identify key factors affecting churn
• Create an interactive Power BI dashboard for business insights

---

Dataset Information

The dataset contains customer information such as:

• Demographics (gender, partner, dependents)
• Account details (tenure, contract type)
• Services (internet, streaming, security)
• Billing (monthly charges, total charges)
• Target variable: Churn (Yes/No)

---

Data Preprocessing

• Handled missing values in TotalCharges
• Converted data types from object → numeric
• Removed irrelevant column (customerID)
• Performed One-Hot Encoding for categorical features
• Converted boolean values to numeric (0/1)

---

Exploratory Data Analysis (EDA)

Key Insights:

• Customers with month-to-month contracts have the highest churn
• Customers with low tenure (new customers) are more likely to churn
• Higher monthly charges increase churn probability
• Customers with long-term contracts are more stable

---

Machine Learning Models

1. Logistic Regression

• Initial model showed good accuracy (~79%)
• Struggled with identifying churn customers (low recall)

2. Improved Logistic Regression

• Applied class_weight='balanced'
• Improved churn detection significantly
• Recall increased from 52% → 79%

3. Support Vector Machine (SVM)

• Achieved highest recall (~83%)
• Best model for detecting churn customers

---

Model Evaluation Metrics

• Accuracy
• Confusion Matrix
• Precision, Recall, F1-score

Important Finding:

Recall is more important than accuracy in churn prediction because missing a churn customer leads to business loss.

---

Feature Importance

Top factors influencing churn:

• Tenure (most important)
• Contract Type
• Monthly Charges
• Internet Service (Fiber Optic)

---

Power BI Dashboard

An interactive dashboard was created to visualize churn insights:

Dashboard Features:

• KPI Cards (Total Customers, Churn Count, Churn Rate)
• Churn Distribution (Pie Chart)
• Churn by Contract Type
• Churn by Tenure Group
• Churn by Charges Group
• Internet Service Impact
• Interactive Filters (Gender, Contract, Service)

---

Business Insights

• Focus on improving early customer experience
• Encourage long-term contracts
• Review pricing for high-paying customers
• Target high-risk customers using predictive models

---

Tools & Technologies Used

• Python (Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn)
• Machine Learning (Logistic Regression, SVM)
• Power BI (Dashboard & Visualization)
• Statistics & Data Analysis

---

Project Structure

├── notebooks/
│   ├── 01_data_cleaning.ipynb
│   ├── 02_model_building.ipynb
│   ├── cleaned_churn.csv
│   ├── churn_cleaned_readable.csv
├── output images/
│   ├── Data cleaning images/
│       ├── Churn VS 1 year Contract
│       ├── Churn VS 2 year Contract
│       ├── Customer churn Distribution chart
│       ├── Monthly charges VS Churn
│       ├── Tenure VS Churn
│   ├── Model building images/
│       ├── Confusion matrix Heatmap
│       ├── Top 10 Important features
├── PowerBI file/
│   ├── churn_dashboard.pbix
│
└── README.md

---

Conclusion

This project demonstrates an end-to-end data analytics workflow:

• Data Cleaning
• Data Analysis
• Machine Learning
• Business Insights
• Dashboard Visualization

It highlights how data-driven approaches can help businesses reduce customer churn and improve decision-making.

---

If you found this project useful, consider giving it a star!