Churn Prediction with MLflow

Project Description

This project demonstrates how to use MLflow to track and manage the machine learning lifecycle for a Customer Churn prediction task. We use the IBM Telco Customer Churn dataset to predict which customers are likely to leave the service.

Module 1: MLflow Introduction

The goal of this module is to establish experimental rigor and learn how to compare models using MLflow Tracking.

Learning Objectives

• Tracking Experiments: Log parameters and metrics during model training and evaluation.
• Artifact Management: Save and organize model performance visualizations (confusion matrix, ROC curves) and the trained model itself.
• MLflow UI Analysis: Navigate the UI to compare different runs and analyze experimental results.
• Model Candidate Identification: Use documented evidence to select the best candidate model.