决策树分类的属性选择方法的研究  

Research on an Attribute Section Method for Decision Tree

作　　者：王会青[1] 陈俊杰[1] 侯晓晶[1] 郭凯[1] 

机构地区：[1]太原理工大学计算机科学与技术学院,太原030024

出　　处：《太原理工大学学报》2011年第4期346-348,352,共4页Journal of Taiyuan University of Technology

基　　金：国家自然科学基金(60970059);国家科技支撑计划基金(2009BAH42B02)

摘　　要：针对ID3算法偏向于选择取值较多的属性作为测试属性的缺点,引入OneR算法选择属性的相关子集进行分类,降低无关属性和重复属性对分类的影响。实验结果表明,与ID3算法相比优化后的方案提高了ID3算法的分类准确率,降低了分类时间,同时克服了ID3算法的取值偏置问题,优化了分类结果。As ID3 algorithm is inclined to choose the attributes with more values and less effect on classification as the test attributes, the OneR algorithm is introduced to select the rele- vant attribute subsets to classify the dataset, reduce the effect of irrelevant and repeated attributes. The experimental results show that the improved solution has higher classification accuracy and less classification time over the ID3 algorithm, overcomes the problem of value deviation, lastly optimizes the classification results.

关 键 词：决策树 ID3算法 属性选择 怀卡托智能分析环境(WEKA)系统 

分 类 号：TD301.6[矿业工程—矿井建设]