Scholarlink

In this paper, we introduce novel voting-based pruning strategies for the aggregation of Random Forests into decision tree-like explainable models. These strategies are designed to mitigate the enormous explosion in size during the aggregation process while minimizing the loss of accuracy. We explore four strategies ranging from strict semantics-preserving methods to more dynamic stochastic and machine learning models, which balance the trade-off between pruning aggressiveness and fidelity to the original Random Forest. We illustrate the impact of our approach with experiments covering an extensive set of popular datasets in this field.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Voting-Based Shortcuts through Random Forests for Obtaining Explainable Models

  • 章节

  • 发表日期：

    2025年

查看全文
  • Alnis Murtovi,
  • Maximilian Schlüter,
  • Bernhard Steffen

摘要

In this paper, we introduce novel voting-based pruning strategies for the aggregation of Random Forests into decision tree-like explainable models. These strategies are designed to mitigate the enormous explosion in size during the aggregation process while minimizing the loss of accuracy. We explore four strategies ranging from strict semantics-preserving methods to more dynamic stochastic and machine learning models, which balance the trade-off between pruning aggressiveness and fidelity to the original Random Forest. We illustrate the impact of our approach with experiments covering an extensive set of popular datasets in this field.