very_simple_fela - Engineer features for Taobao and Dia AUC benchmarks
Provides dataset-specific feature engineering guidance and code patterns for improving AUC on Taobao or Dia tabular benchmarks through iterative evaluation.
Tags
Updated: 2026-09-28Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Detect the active dataset
- Inspect benchmark schema
- Engineer dataset-specific features
- Modify train and test features
- Evaluate feature sets with AUC
- Apply evaluation early stopping
- Prevent target leakage
- Compare feature subsets
Inputs
- Benchmark work directory
- Dataset identifier
- Early-stop patience
- Benchmark feature schema
- Taobao history logs
- Dia train and test columns
Outputs
- AUC evaluation scores
- Updated training feature table
- Updated test feature table
- Feature selection state
- Early-stop status
Requirements
- Python runtime
- Benchmark API package
- Access to benchmarks/feature_engineering
- Permission to modify benchmark tables