Customer Churn Prediction with Vertex AI
This project implements an end-to-end machine learning pipeline for predicting customer churn using Google Cloud's Vertex AI and AutoML. The model analyzes 12 customer features including tenure, usage patterns, support interactions, and payment behavior to predict churn probability. The dataset contains 440,833 customer records with both numerical and categorical features. After exploratory data analysis and feature engineering, the data is uploaded to Google Cloud Storage and used to train an AutoML classification model. The deployed model achieves 99.9% prediction confidence and provides real-time inference through a Vertex AI endpoint, enabling businesses to identify at-risk customers and take proactive retention actions.
Key Metrics
Highlights
- AutoML model training with automatic hyperparameter tuning
- Real-time prediction endpoint on Vertex AI
- Feature importance analysis with Shapley values
- 440K+ customer records for robust training
This is a notebook-based ML project. View the full implementation on GitHub.
Features
- AutoML model training with automatic hyperparameter tuning
- Real-time prediction endpoint on Vertex AI
- Feature importance analysis with Shapley values
- 440K+ customer records for robust training
- 99.9% prediction confidence on test data
- Scalable cloud-native deployment
Architecture
┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Dataset │────▶│ Cloud │────▶│ AutoML │
│ (CSV) │ │ Storage │ │ Training │
└─────────────┘ └─────────────┘ └─────────────┘
│
▼
┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Client │◀────│ Vertex │◀────│ Model │
│ Request │ │ Endpoint │ │ Registry │
└─────────────┘ └─────────────┘ └─────────────┘Tech Stack
Key Learnings
AutoML significantly reduces time-to-model while maintaining high accuracy
Feature engineering on customer behavior data is crucial — support calls and payment delays are strong predictors
Vertex AI's managed endpoints simplify deployment but require careful cost monitoring
Shapley values provide interpretable explanations essential for business stakeholders
Want to see more AI projects?
Check out the rest of my AI Lab or get in touch to discuss AI/ML collaboration.