An framework for generating fuzzy decision trees based on fuzzy rough techniques

Fuzzy decision tree is useful for expressing fuzzy knowledge because of its readability. There have been several induction algorithms for fuzzy decision trees from real value data sets. In this paper, we propose a framework for generating fuzzy decision trees based on fuzzy rough techniques. Firstly, the ordinary fuzzification techniques are replaced by a clustering technique based on the tolerance relation corresponding to every attribute. Secondly, a new fuzzy rough technique is introduced to reduce the dimensions. Thirdly, a new heuristics employing fuzzy rough lower approximation is constructed to generate a fuzzy decision tree. A small data is used for demonstrating the practicability of the proposed method.