New LMDI+ method enhances interpretability for tree-based models

Researchers have developed Local MDI+ (LMDI+), a new method for quantifying feature importances in tree-based models for individual samples. Unlike existing approximation-based methods, LMDI+ leverages the internal structure of decision trees and linear models. Across twelve benchmark datasets, LMDI+ demonstrated a 10% improvement in predictive performance when using only selected features and showed greater stability in feature importance rankings. The method also proved effective in identifying counterfactuals and discovering subgroups in a housing dataset case study. AI

IMPACT Enhances the interpretability of widely used tree-based models, potentially increasing trust and adoption in high-stakes applications.

RANK_REASON This is a research paper detailing a new method for feature importance in machine learning models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• LMDI+
• Local MDI+
• MDI+
• TreeSHAP

• paper
• other

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