论文部分内容阅读
为提高传统决策树学习方法的扩展性和自适应性,基于广义信息论提出决策森林多重子模型集成方法.采用从下至顶的学习策略,将离散化处理和决策树的逻辑表达有机结合在一起,整个学习过程不需要任何人为参与,能自动确定子树数目和子树结构.在UCI机器学习数据集上的实验结果和样例分析验证了本文方法的可行性和有效性.
In order to improve the scalability and adaptability of the traditional decision tree learning method, a multi-sub-model approach to forest decision-making decision-making is proposed based on the generalized information theory.Using the learning strategy from the bottom to the top, the logic of discretization and decision tree is organically combined, The entire learning process does not require any human involvement, and can automatically determine the number of sub-trees and sub-tree structure.The experimental results and sample analysis on the UCI machine learning data set verify the feasibility and effectiveness of the proposed method.