Extraction of quantified fuzzy rules from numerical data

We propose a method to extract quantified fuzzy rules from numerical data. An example of this type of fuzzy rule is “Most data whose attribute A is large are small in the attribute B”, where the “large” and “small” are fuzzy sets of attributes A and B, respectively, and “most” as a fuzzy quantifier. For selecting a combination of fuzzy sets in attributes for fuzzy rules, we use a fuzzy ID3-based method to generate a fuzzy decision tree for a specified class. From each tree, we extract a quantified fuzzy rule from a path of the root to a class node by evaluating its understandability and informativeness. We apply the method to Iris classification problem by Fisher (1936) and diagnosis data by gas in oil