期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2015
卷号：3
期号：10
DOI：10.15680/IJIRCCE.2015.0310068
出版社：S&S Publications
摘要：One of the simplest and popular classification algorithms is the decision tree. While classification algorithms use a target attribute, Clustering algorithms group the data without a target attribute. K-Means is the simplest of the clustering algorithms. In a decision tree, branching of a nominal, ordinal, discrete or binary attribute is simple and straight forward compared to branching of a continuous attribute which is trickier, commonly creating a binary branch. Here, we present a hybrid algorithm that combines decision tree and K-Means to create multiple branches for a continuous attribute at a node. For the 9 standard datasets tested from UCI repository, the HDTKM algorithm gives an average accuracy of 78.1% and J48 gives an average accuracy o f 76.1%.
关键词：Decision tree; K -Means; C4.5; Continuous data
