Customer Churn Prediction

Problem

Customer churn significantly impacts recurring revenue. This project builds a machine learning pipeline to predict high-risk customers and simulate retention targeting strategies

Dataset

• 440,00+ customer records
• Features include: Tenure, Usage Frequency, Support Calls, Payment Delay, etc.
• Bianry target variable: Churn (1 = churned, 0 = retained)

Project Structure

• api/
  • app.py
• notebooks/
• eda.ipynb
• src/
  • preprocessing.py
  • train.py
  • evaluate.py
• models/
  • churn-prediction.joblib
• main.py
• requirements.txt
• README.md

Methodology

The project follows a standard machine learning pipeline:

1. Data Loading
2. Data Cleaning
3. Exploratory Data Analysis (EDA)
4. Feature Preprocessing
5. Model Training
6. Model Evaluation
7. Model deployment via FastAPI

Model comparison in progress. Planned models include:

• Logistic Regression (baseline)
• Random Forest
• Gradient Boosting

Evaluation will focus on:

• ROC-AUC
• Precision / Recall
• Confusion Matrix

Results

To be finalised after model benchmarking and evaluation

Business Implications

Insights will be dreived after selecting the optimal recall-precisin balance for churn detection.

How to Run

1. Create virtual environment python -m venv venv

2. Activate environment source venv/bin/activate or Windows equivalent

3. Install dependencies pip install -r requirements.txt

4. Run the pipeline python main.py

Model Performance (Logsitic Regression)

ROC Curve

The model achieved an ROC-AUC score of 0.91, indicating strong classification performance.

Classification Report

Confusion Matrix

Model Performance (Random Forest)

ROC Curve

The model achieved an ROC-AUC score of 0.95, indicating a stronger classification performance.

Classification Report

Confusion Matrix

TO DO

Model Performance (Gradient Boost)