A deep learning based neuro-fuzzy approach for solving classification problems

• Main Authors: Talpur, N., Abdulkadir, S.J., Hasan, M.H.
• 格式: Conference or Workshop Item
• 出版: Institute of Electrical and Electronics Engineers Inc. 2020
• 在線閱讀: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85097567980&doi=10.1109%2fICCI51257.2020.9247639&partnerID=40&md5=e789247d0e8c1c0da02ec37a185a5ca0; http://eprints.utp.edu.my/29876/

Techniques involved artificial intelligence and machine learning offers various classification methods in order to deal with daily life problems. Among these methods, Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and Deep Neural Network (DNN) are the most commonly used classifiers. Since ANFIS is not suitable for high-dimensional data, therefore DNN was introduced to overcome this problem faced by conventional methods. However, due to the optimization of millions of parameters in their deep architecture, the decision made by DNN faced the criticism of being non-transparent. To overcome this problem, recently, various researchers are coming up with the idea of using fuzzy logic with DNN. Therefore, this study also proposed a Deep Neuro-Fuzzy Classifier (DNFC) with a cooperative based structure for solving classification problems, particularly. The performance of the proposed DNFC was evaluated with ANFIS and DNN classifier, where overall results show that the performance of ANFIS classifier decreased when input size increased. While the performance of the proposed model demonstrated nearly similar or slightly higher accuracy as compared to DNN. © 2020 IEEE.