基于PCA的决策树优化算法  

PCA-based Decision Tree Optimization Algorithm

作　　者：谢霖铨[1] 徐浩 陈希邦 赵楠 XIE Lin-quan;XU Hao;CHEN Xi-bang;ZHAO Nan(College of Science,Jiangxi University of Science and Technology,Ganzhou 341000,China)

机构地区：[1]江西理工大学理学院

出　　处：《软件导刊》2019年第9期69-71,76,共4页Software Guide

基　　金：国家自然科学基金项目(61762047);国家重点研发计划重点专项项目(2016YFB0800700)

摘　　要：为了改善传统ID3算法在分类属性选择上存在多值偏向性的不足,提出基于PCA的决策树优化算法。在普通基于PCA的决策树改进算法中,存在数据经降维处理后代表性不强的问题,导致算法需经过多次数据运行后,准确率才能小幅提升。在ID3算法基础上,在分类前两次提取属性特征值,并计算了需要分类的数据量,也即对原始数据进行最重要的属性选择。在子树建立之后,再进行数据的降维合并选择。采用UCI数据库中的3个数据集对改进算法进行验证,结果表明改进算法的平均准确率达到94.6%,相比传统ID3算法与普通PCA决策树优化算法分别提升了1.6%和0.6%。因此,基于PCA的决策树算法能在一定程度上提升结果准确率,具备一定的应用价值。In this paper, the problem of the multi-valued bias of the traditional ID3 algorithm in classification attribute selection is improved. A PCA-based decision tree optimization algorithm is proposed. In the ordinary PCA-based decision tree improvement algorithm, there are data after dimension reduction processing. The problem of low representation is that the improved algorithm needs to pass through multiple data to bring the accuracy to increase slightly. Therefore, based on the ID3 algorithm, the feature values are extracted twice before classification, and the classification needs to be calculated. The amount of data, that is, the most important attribute selection for the original data, after the subtree is established, the data is reduced and merged and selected. In the experimental stage, the improved algorithm was verified by three data sets in the UCI database. The results showed that the average accuracy rate in the three data sets reached 94.6%, and the traditional ID3 algorithm and the ordinary PCA decision tree optimization algorithm were improved by 1.6% and 0.6%, which proves the algorithm has certain practical significance.

关 键 词：决策树算法 ID3 PCA算法 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]