基于决策树规则的分类算法研究  

Classification algorithm of rule based on decision-tree

作　　者：谭俊璐[1] 武建华[2] 

机构地区：[1]暨南大学信息科学技术学院计算机系,广东广州510632 [2]暨南大学珠海学院计算机科学系,广东珠海519070

出　　处：《计算机工程与设计》2010年第5期1017-1019,共3页Computer Engineering and Design

基　　金：暨南大学青年基金项目(51208030)

摘　　要：在商业利益的驱动下,人们不断地深入研究决策树算法。为了提高分类的精度,提出了一种基于决策树规则的分类算法。通过C4.5决策树算法得出决策规则,计算决策规则的长度、准确率与覆盖率,对所得的决策规则依次按照规则长度与准确率的乘积大小、长度的大小、覆盖率的大小对规则集进行排序构造分类器,选择优选权最高的规则进行匹配分类。实验结果表明,与C4.5算法相比,该方法的分类精度有所提高。Driven by the business advantage, people continuously lucubrate decision-tree algorithm to improve the precision of classification. For this purpose, a decision-tree rule classification algorithm is proposed. The decision rule are gotten by C4.5 decision tree, then the length of decision rules, the accuracy and the coverage are calculated, then the corresponding decision rule ranking approach： the priority of the value of the length of rule multiplies by the accuracy of rule is first, the length of the rule is second, and the coverage of the rule is thirdly, then classifying through matching rules one by one. The testing results show that the proposed method has abetter precision of classification than the C4.5 methods.

关 键 词：决策树 分类算法 C4.5决策规则分类 规则排序策略 规则质量 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]