A Decision Tree Algorithm Based on Dispersion Measure of Attribute Information

Abstract

In this paper, an improved decision tree algorithm based on dispersion measure of attribute information was proposed, which combined information gain and dispersion of attribute information as an evaluation criterion of attribute selection in order to overcome the deficiency that ID3 decision tree algorithm leaned to the multi-value attribute. From results of the experiment, it can be demonstrated that the proposed algorithm could over the deficiency of leaning to the multi-value attribute, and has good performance on classification.