关键词: 决策树, 数值型属性, 聚类区间, 分类

Abstract: This paper proposes such an algorithm，called CAMNA（Clustering and Merging on Numerical Attributes），which is a new algorithm of unsupervised discretization based on clustering.The method divides a set of the numerical attribute values into many intervals based on clustering in the first step.Then in the second step，the cluster quality is optimized by computing the class label of the adjacent intervals.This procedure can not stop until a satisfactory discretization schema is reached.Experimental evaluation of several discretization algorithms show that the proposed algorithm is more efficient and can generate a better discretization schema.Comparing the output of C4.5，resulting tree is smaller，less classification rules，and high accuracy of classification.

Key words: decision tree, numerical attributes, clustering intervals, classification