改进的有序决策归纳算法.pdfVIP

Classified Index: CODE: 10075 U.D.C: NO: A Dissertation for the Degree of M.Engineering Improved Ordinal Decision Tree Induction Algorithms Candidate: Pan Pan Supervisor: Prof. Wang Xizhao Academic Degree Applied for: Master of engineering Specialty: Computer Application Technology University: Hebei University Date of Oral Examination: May , 2014 摘 要 摘 要 在已有的有序决策树归纳算法中,选择条件属性和决策属性的有序互信息值最大的 属性作为扩展属性,没有考虑到那些取值多且有序的条件属性和属性之间的相关性,这 就造成了重复选择某个条件属性为扩展属性,这样会使得分类器的分类效果不佳，测试 集的精确度不高。针对这一问题，本文对有序决策树归纳算法进行了改进，主要工作包 括以下两个方面： 1．提出了一种新的有序决策树归纳算法。该算法选择的扩展属性不仅和类的有序 互信息值最大，而且要求和同一分支上已被用过的条件属性的有序互信息值最小，可避 免同一条件属性的重复选择,真正体现了条件属性和决策属性之间的有序互信息,从而 提高了测试精度。 2 ．在有序数据集中，有的属性取值多且有序，也会影响扩展属性的选择，使测试 精度降低，考虑到这个问题，用有序互信息率代替有序互信息来选择扩展属性，克服选 择扩展属性时偏向取值多且有序的属性，从而提高测试精度。 关键词 决策树 有序分类 有序互信息 有序信息熵 有序互信息率 I Abstract Abstract In the existing methods for inducing ordinal decision tree, ranking mutual information between the conditional attributes and the decision attribute is employed as a heuristic to select the extended attribute, while more and ordinal values of attribute and correlation among the conditional attributes are not considered, this has resulted in repeating an attribute for extended attributes. This will make classifier ineffective and lower testing accuracy. In order to deal with this problem, the ordinal decision tree induction algorithm is improved in