A Novel and Efficient Neuro-Fuzzy Classifier for Medical Diagnosis

This study attempts to propose a novel neuro-fuzzy network which can efficiently reason fuzzy rules based on training data to solve the medical diagnosis problems. First, this study proposes a refined k-means clustering algorithm and a gradient-based learning rules to logically determine and adaptively tuned the fuzzy membership functions for the employed neuro-fuzzy network. In the meanwhile, this study also presents a feature reduction scheme based on the grey-relational analysis to simplify the fuzzy rules obtained from the employed neuro-fuzzy network. Experimental results indicated that the proposed neuro-fuzzy network with feature reduction can discover very simplified and easily interpretable fuzzy rules to support medical diagnosis.