Enhance Neuro-fuzzy system for classification using dynamic clustering

Abstract

The Enhance Neuro-fuzzy system for classification using dynamic clustering presents in this paper is an extension of the original Neuro-fuzzy method for linguistic feature selection and rule-based classification. The new algorithm resolves the limitations of the original algorithm that uses only 3 membership functions for all features to fine the appropriate function for each feature. Each feature of the dataset is pre-processed by a new approach to clustering automatically. The Neuro-fuzzy classification models for each dataset is created in accordance with the number of clusters have been divided for each feature. In order to be appropriate functioning in the Neuro-fuzzy structure, a new algorithm has been adapted to use the binary instead of the bipolar as original algorithm. Thirteen datasets were used to test the performance of the proposed algorithm. The average accuracy calculated from the 10-fold cross validation found that this method can increase performance of the already proof high accuracy Neuro-fuzzy for classification.