Fetch Point-in-Time Correct Features
Avoid label leakage when building training sets.
The Time-Travel Problem
To train fairly, each label needs the feature values that were known before the event, not after. Mixing up time leaks the future into your model. ⏳
What Label Leakage Is
Label leakage happens when a feature secretly carries information from after the prediction moment. The model looks brilliant offline, then flops live.
All lessons in this course
- Why Feature Stores Exist
- Define Feature Views with Feast
- Materialize Features to an Online Store
- Fetch Point-in-Time Correct Features