基于相似依赖度的属性加权决策树算法

摘要

Rough set can process on existing knowledge without any prior knowledge using data itself. It can simplify and get property combination on premise of critical information. On basis of this theory a feature selection algorithm based on relative dependence is proposed. It mainly choose the most similar property core to decision property from reduced properties. The most similar property combination is used to data mining for decision tree. The proposed algorithm is efficient and feasible by analysising these data, which is factors affecting students grades. And the veracity of classification is also improved.%粗糙集分析方法利用数据本身提供的信息,不需要任何先验知识即可对已有的知识进行处理,在保留关键信息的前提下,对数据进行简化并求得属性组合.在此基础上提出了一种基于相似依赖度的属性选择算法,从约简出的属性组合中选择与决策属性最为相似的核集,根据属性的相似依赖度作为决策树的加权值,从而建立决策树.通过对影响学生成绩因素的数据进行分析表明,所提出的算法是易于实施的,而且形成的决策树的准确率也有了一定的提高.