基于最大决策熵的快速属性约简算法  

Fast Attribute Reduction Algorithm Based on Maximum Decision Entropy

作　　者：袁梅 

机构地区：[1]烟台大学计算机与控制工程学院,山东 烟台

出　　处：《数据挖掘》2023年第3期222-229,共8页Hans Journal of Data Mining

摘　　要：在大数据时代背景下,各领域数据爆炸式增长,数据类型复杂多样。针对决策系统中基于最大决策熵的属性约简算法在大规模数据集下运行效率低的问题,提出了一种基于启发式的快速属性约简算法。本文提出的算法首先研究了属性和对象在属性约简过程中的变化对其产生影响,其次提出了属性重要度保序性的相关定理。最后通过UCI数据集对提出算法的有效性进行验证,结果表明提出的快速属性约简算法的运行效率更高。In the era of big data, data in various fields is growing explosively, and data types are complex and diverse. Aiming at the low efficiency of attribute reduction algorithm based on maximum decision entropy in decision system under large data sets, a fast attribute reduction algorithm based on heuristic is proposed. The algorithm proposed in this paper firstly studies the influence of the changes of attributes and objects in the process of attribute reduction, and then puts forward the related theorem about the rank preservation of attributes. Finally, the effectiveness of the proposed algorithm is verified by the UCI data set, and the results show that the proposed fast attribute re-duction algorithm is more efficient.

关 键 词：属性重要度 属性约简算法 大规模数据集 决策熵 数据类型 决策系统 算法的有效性 保序性 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]