Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·Aug 31·5 min read
Timing-Aware Repurchase Prediction for Web-Scale E-Commerce: Survival Models for Multi-Surface Grocery Recommendation
TL;DR
Survival models for predicting repurchase timing in e-commerce outperform traditional binary classifiers by directly estimating time-to-repurchase, achieving significant efficiency and accuracy improvements.
✦ Why It Matters
Consider adopting survival models for your e-commerce recommendation systems to enhance prediction accuracy and efficiency.
Key Takeaways
- Implement survival models to predict repurchase timing directly.
- Use Exponential AFT for applications requiring high probability accuracy.
- Leverage Log-Normal for ranking-focused recommendations.
- Analyze feature importance shifts when transitioning to survival models.
How It Works
Survival models estimate the time until an event occurs, in this case, a repurchase. By using an Accelerated Failure Time (AFT) framework, the model can predict repurchase timing without needing separate classifiers for different time horizons.
This approach allows for a more efficient use of resources, as fewer trees are required to achieve comparable or better performance.