计算机应用研究2018，Vol.35Issue(4)：1018-1022,5.DOI:10.3969/j.issn.1001-3695.2018.04.013
基于信息熵和几何轮廓相似度的多变量决策树
Multivariate decision tree based on information entropy and outline similarity
摘要
Abstract
The existing multivariate decision tree is better than the univariate decision tree in the aspect of classification accuracy and tree structure complexity,but its training time complexity is higher than the univariate decision tree,so the existing multivariate decision tree does not apply to classification tasks which have fast response.Due to the problem of high training time which the multivariable decision tree has,this paper proposed a new multivariate decision tree algorithm:a multivariate decision tree based on information entropy(IEMDT).IEMDT projected a n-dimension data point on a one-dimension line by using the specification of one to one mapping which geometric outline similarity function has,thus received an ordered projection points,then IEMDT searched the best splited point collection through class projection boundary and information entropy,which splited projection point collection into several subsets,and continued to project and split the corresponding sub datasets.Finally it generated the decision tree.The experimental results show that IEMDT has lower training time,but also has higher classification accuracy.关键词
多变量决策树/分类/单变量决策树/几何轮廓相似度/信息增益Key words
multivariate decision tree/classification/univariate decision tree/outline similarity/information gain分类
信息技术与安全科学引用本文复制引用
张宇,包研科,邵良杉..基于信息熵和几何轮廓相似度的多变量决策树[J].计算机应用研究,2018,35(4):1018-1022,5.基金项目
国家自然科学基金资助项目(71371091) （71371091）
