A self learning rough fuzzy neural network classifier for mining temporal patterns

Abstract

This paper proposes a new approach that integrates neural networks with the fuzzy rough set to build a Rough Fuzzy Neural Network Classifier (RFNNC) in order to mine temporal patterns in clinical databases. The lower approximation hypothesis and fuzzy decision table with the fuzzy features are used to acquire the fuzzy decision classes for deciding on the attributes. By contemplating a subset of attributes, comprising of the temporal intervals, the lower approximations are devised in this work. Moreover the basic sets are attained from lower approximations are sorted into the decision classes. The discernibility of the decision classes is designed to delineate the temporal consistency degree between the objects of the sets, from which the reducts are acquired. Next, the attribute subset from the reducts is used for training the fuzzy neural network to infer fuzzy rules. The induced rules will result with temporal patterns for classification. The fuzzy neural network has completely used the competence of fuzzy rough set theory to condense huge quantity of superfluous data. The effectiveness of this method is compared with other classifiers such as fuzzy rule based classifier to evaluate the accuracy of the proposed fuzzy neural network classifier. Experiments have been performed on the diabetic dataset and the simulation results induced proves that the proposed fuzzy neural network classifier on medical diabetic dataset stays as a corroboration for predicting the severity of the disease and exactness in decision support system.