基于决策树学习中的测试生成及连续属性的离散化
TEST GENERATION AND DISCRETIZATION OF CONTINUOUSLY VALUED ATTRIBUTES IN DECISION TREE BASED LEARNING
-
摘要: 文中介绍并分析了基于决策树学习中的测试评价标准、测试生成机制及连续型属性的离散化等方法和实现技术.通过分析表明,在离散化过程中,采用信息熵最小化启发式能带来较好的效果.与二分离散化方法相比,采用多分离散化方法能从相同的实例集中构造出更好的决策树.Abstract: Here introduced and analyzed are the methods of test evaluation criterion,test generation,and discretization of continuously valued attributes.It can be seen through the analysis that the heuristics of minimizing information entropy show good results.Compared with the bi interval discretization method,the multi interval discretization method can produce better decision trees.
下载: