摘要
应用粗糙集理论,提出了一种利用新的启发式函数构造决策树的方法。该方法以属性重要性评价指标作为信息熵函数,对条件属性进行选择,充分考虑了属性间的依赖性和冗余性,弥补了ID3算法对属性间依赖性强调不够的缺点,解决了决策树中子树的重复和有些属性在同一决策树上被多次选择的问题,该方法还能对不相容决策表进行正确分类。实例表明该方法是正确有效的,而且明显优于传统的决策树构造方法。
A method used a new heuristic function based on rough sets was proposed to build a decision tree. In view of shortcomings existing in IDa algorithm, the method regards the evaluating of significant attributes as the information entropy to select the condition attribute, and the dependability and redundancy between attributes are taken into account sufficiently. So the repetition of the decision subtrees and some attributes to be chosen many times on the same decision tree were resolved, and consistent tables and inconsistent tables can be classified correctly. The example showed that the method is better than the traditional method and has been verified to be effective.
出处
《南京工业大学学报(自然科学版)》
CAS
2005年第4期80-83,共4页
Journal of Nanjing Tech University(Natural Science Edition)
关键词
粗糙集
决策树
属性约简
rough sets
decision tree
attributes reduction