Automatic generation of fuzzy classification rules using granulation-based adaptive clustering

A central problem of fuzzy modelling is the generation of fuzzy rules that fit the data to the highest possible extent. In this study, we present a method for automatic generation of fuzzy rules from data. The main advantage of the proposed method is its ability to perform data clustering without the requirement of predefining any parameters including number of clusters. The proposed method creates data clusters at different levels of granulation and selects the best clustering results based on some measures. The proposed method involves merging clusters into new clusters that have a coarser granulation. To evaluate performance of the proposed method, three different datasets are used to compare performance of the proposed method to other classifiers: SVM classifier, FCM fuzzy classifier, subtractive clustering fuzzy classifier. Results show that the proposed method has better classification results than other classifiers for all the datasets used.