New qshap tool decomposes R-squared for gradient-boosted trees

Researchers have introduced qshap, a new method for decomposing R-squared values in gradient-boosted decision trees. This tool, available in R and Python, quantifies the contribution of individual features to overall model performance by analyzing the quadratic loss of observations. qshap supports popular GBDT implementations like XGBoost, LightGBM, and Catboost, utilizing efficient C++ backends and offering specialized acceleration for oblivious trees. AI

影响 Provides a new method for understanding feature importance in gradient-boosted trees, potentially improving model interpretability and debugging.

排序理由 The cluster describes a new method and tool for analyzing machine learning models, presented in an academic paper. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv stat.ML 阅读 →

• Catboost
• CPP
• LightGBM
• Python
• qshap
• R
• Shapley Values
• XGBoost

• 论文
• 模型发布

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →