FDT 1.0: An improved fuzzy decision tree induction tool

Abstract

FDT is a scalable supervised-classification freeware software tool implementing fuzzy decision trees. It is based on an improved version of the fuzzy ID3 (FID3) algorithm. It implements four different variations of FID3, the first utilizes fuzzy information gain, the second utilizes classification ambiguity, the third utilizes a fuzzy version of Gini-index and the fourth integrates fuzzy information gain and classification ambiguity to select a test (branching) feature. FDT also implements our previously published rule-set reduction method. The tool supports two inference methods: sum-of-products (X-X-+) and max-min. In this paper we introduce FDT and review its' major features and functionalities. In addition, we show that integrating our previously published rule-set reduction approach can improve the classification accuracy and can reduce the number of rules produced of all FID3 versions.