A Fuzzy Decision Tree Algorithm Based on C4.5

Abstract

Decision trees have been successfully applied to many areas for tasks such as classification, regression, and feature subset selection. Decision trees are popular models in machine learning due to the fact that they produce graphical models, as well as text rules, that end users can easily understand. Moreover, their induction process is usually fast, requiring low computational resources. Fuzzy systems, on the other hand, provide mechanisms to handle imprecision and uncertainty in data, based on the fuzzy logic and fuzzy sets theory. The combination of fuzzy systems and decision trees has produced fuzzy decision tree models, which benefit from both techniques to provide simple, accurate, and highly interpretable models at low computational costs. In this paper, we expand previous experiments and present more details of the FuzzyDT algorithm, a fuzzy decision tree based on the classic C4.5 decision tree algorithm. Experiments were carried out using 16 datasets comparing FuzzyDT with C4.5. This paper also includes a comparison of some relevant issues regarding the classic and fuzzy models.