New TOP-Gain method improves random forest feature selection accuracy

Researchers have introduced TOP-Gain, a novel criterion for random forest wrapper feature selection designed to mitigate noise in out-of-bag predictions. This method, which truncates decreases in probability gain at zero, demonstrated superior performance in a benchmark study across 26 OpenML-CC18 datasets. TOP-Gain achieved the best mean rank for ROC-AUC, accuracy, and Macro-F1, outperforming 13 of 14 baselines while significantly reducing computational cost compared to VSURF. AI

IMPACT Enhances the efficiency and robustness of feature selection in machine learning models.

RANK_REASON The cluster describes a new method for feature selection in random forests, detailed in a research paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Hugging Face Daily Papers →

• Accuracy
• Macro F1
• OpenML CC18
• random forest
• ROC-AUC
• TOP-Gain
• Truncated Out-of-bag Probability Gain
• VSURF

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →