Fuzzy Clustering-Based Neural Fuzzy Network with Support Vector Regression
Chia‐Feng Juang, Cheng-Da Hsieh, Jyun-Lang Hong · 2010
This paper proposes a new fuzzy regression model, the Fuzzy Clustering-based Fuzzy Neural Network with Support Vector Regression (FCFNN-SVR). Structurally, a FCFNN-SVR is a five-layered network. The consequent layer in FCFNN-SVR is of Takagi-Sugeno (TS)-type consequent, which is a linear function of system inputs. For structure learning, a one-pass clustering algorithm clusters the input training data and determines the number of network nodes in hidden layers. For parameter learning, a linear support vector regression (SVR) algorithm is proposed to tune free parameters in the consequent part. The motivation for using SVR for parameter learning is to improve the FCFNN-SVR generalization ability. This paper demonstrates the capabilities of FCFNN-SVR by conducting simulations in clean and noisy function approximations. This paper also compares simulation results from the FCFNN-SVR with Gaussian kernel-based SVR and other learning models.