Neural network fuzzy control based on clustering analysis

The neural network-fuzzy controller is more improved than previous controller. But there are some deficiencies, such as confirming number of fuzzy subset, picking up fuzzy rules and confirming the initial value of membership. Aiming at these deficiencies, the thesis brings up the neural network-fuzzy control based on clustering analysis. The number of fuzzy subset can be confirmed by means of confirming number of best clustering in which the valid function of clustering is used. Confirming initial value of membership function and getting rules can be carried through by means of clustering analysis. The neural network-fuzzy controller based on clustering analysis can be gained after training of neural network clustering analysis is a method of data disposing and include hard clustering, fuzzy clustering and feasibility clustering. The operation speed of hard clustering is higher, er, but it dissevers the relations of samples. Fuzzy clustering overcomes it, but there is local best. The thesis brings up an improved method to overcome it by combining hard clustering and fuzzy clustering.