Multivariate Decision Trees Based on Regression and Discriminant Analysis

Nowadays most researches on decision tree are based on univariate and only few are based on multivariate. Even these researches on multivariate always ignore attributes´ internal relations or get inefficient results. This paper introduces regression analysis and discriminant analysis method into the decision tree construction algorithm so that we can find out the relations among attributes, promote the classification ability of decision trees, and solve the partition problem of continuous-value attributes.