Churn Prediction Pipeline

ML pipeline for customer churn prediction using LightGBM, Prefect, and Evidently.

Demo Video Drift Detection Monitoring Demo

Quick Start

# 1. Copy environment file
make copy-env

# 2. Start services
make up

# 3. Access UIs
# Prefect: http://localhost:4200
# Evidently: http://localhost:7000

Stop Services

make down

Local Development

# Install dependencies
uv sync

# Copy .env
make copy-env
# run prefect server
uv run prefect server start
# Run flow
uv run src/flow/batch_inference.py

Architecture

Service	Port
Prefect Server	4200
Evidently	7000
PostgreSQL	5432
Redis	6379

Make Commands

make copy-env          # Setup .env
make build             # Build Docker image
make up                # Start services
make up-build-debug    # Start with build logs
make start/stop        # Control services
make down              # Stop & remove

Model

• Input: Age, Annual Income, Gender, Membership Duration, Location
• Output: Churn prediction (0/1)
• Algorithm: LightGBM (200 estimators)