#feature-engineering

Automating Feature Preprocessing Beats Manual Tuning for Tabular ML

Study of 15 search algorithms on 45 datasets reveals evolution and random search outperform complex surrogate models for automated feature pipeline construction.