A dynamic neuro-fuzzy approach for pattern classification

Nowadays, the application of neuro-fuzzy methods has been discovered more than ever for pattern recognition. These powerful tools are able to model the reality of data structure as it should be because, in the real world, datasets are defined in a fuzzy concept. In this research, we present a novel neuro-fuzzy method called Fuzzy Growing Map (FGM), combining the dynamic properties of the Growing Self-Organizing Map (GSOM) and fuzzy set theory. FGM is a dynamic neural fuzzy inference system based on if-then rules, which has the ability to generate fuzzy rules based on certain criteria during the learning phase. This approach can be used as a classifier and approximator. In addition, the trained FGM was used to visualize the fuzzy sets as a map, and the structure of the data can easily be revealed in the feature space. To investigate the effectiveness of FGM, several benchmark datasets were analyzed, and the experimental results for classification show improvements in terms of accuracy and topographic error compared to classification algorithms Fuzzy Self-Organizing Map (FSOM)and Counter Propagation Neural Networks (CPNN).

کلیدواژه ها:

Pattern recognition ، Growing Self-Organizing Map ، Neuro-Fuzzy methods ، fuzzy sets ، classification