利用集成剪枝和多目标优化算法的 随机森林可解释增强模型.

Random forest is a classic black-box model that is widely used in various fields. The structural characteristics of black-box models lead to weak model interpretability, which can be optimized with the help of interpretable techniques to promote the application and development of random forest in scenarios with high reliability requirements. This paper constructed a rule extraction model based on ensemble pruning and multi-objective evolutionary algorithm. Ensemble pruning is an effective method for solving the problem of extracting rules from tree models that tend to fall into local optima, and multi-objective evolutionary has several applications in balancing rule accuracy and interpretability. This paper found that it improved interpretability without sacrificing accuracy. This study integrated ensemble pruning technique with a multi-objective evolutionary algorithm, which enhances the interpretability of random forests and helps promote the decision-making application of this model in areas with high interpretability requirements. [ABSTRACT FROM AUTHOR]