Fuzzy classification trees as environmental indicators

Classification trees are valuable data mining tools, but sometimes they lack generality and flexibility. The idea described in this paper is to increase the generality of classification trees by deriving a fuzzy inferential engine from its hierarchical structure. The paper describes the steps through which the tree structure is decomposed and translated into fuzzy rules. To obtain a complete inferential system fuzzy memberships are then added and optimized. The combined algorithm is demonstrated with classification examples and the improvements with respect to the original tree are discussed.