Induction of bi-branches decision tree with fuzzy number-value attribute

Abstract

This paper presents an algorithm regarding the fuzzy number-valued attribute using the information entropy minimization heuristic. The algorithm gives us a desirable behavior of the information entropy of partitioning. The efficiency of the learning algorithm is improved by analyzing the non-stable cut point and the experiment result shows that the number of leaves in decision tree generation is reduced with the raising of level /spl alpha/. Thus, the scale of decision tree and the recognition rate of classification using the proposed algorithm are improved with the raising of level /spl alpha/. To the unknown-classified sample data, the algorithm offers a rapid matching speed. Finally, the example on medical records that we collected in a hospital shows the utility of the proposed algorithm.