Inducing Fuzzy Decision Trees in Non-Deterministic Domains using CHAID

Most decision tree induction methods used for extracting knowledge in classification problems are unable to deal with uncertainties embedded within the data, associated with human thinking and perception. This paper describes the development of a novel tree induction algorithm which improves the classification accuracy of decision tree induction in non-deterministic domains. The research involved applies the principles of fuzzy theory to the CHAID (Chi-Square Automatic Interaction Detection) algorithm in order to soften the sharp decision boundaries which are inherent in traditional decision tree algorithms. CHAID is a decision tree induction algorithm with the main feature of significance testing at each level, leading to the production of trees which require no pruning. The application of fuzzy logic to CHAID decision trees can represent classification knowledge more naturally and inline with human thinking and are more robust when it comes to handling imprecise, missing or conflicting information. The results of applying fuzzy logic to CHAID induced decision trees are presented in this paper. These have been obtained from sets of real world data, and show that the new fuzzy inference algorithm improves the accuracy over crisp CHAID trees. The results show that the increase in performance is dependant upon the inference technique employed and the amount of fuzzification applied.