Research Of Hybrid Learning Algorithm Based On Interval Type-2 TSK Fuzzy Logic System

In recent years, as an important branch of type-2 fuzzy logic systems, type-2 TSK fuzzy logic systems have been applied in system modeling and control of complex nonlinear multivariable. So researching the design algorithm problems of type-2 TSK fuzzy logic systems has practical and realistic significance. At present, type-2 TSK fuzzy logic systems exist the implementation difficulties. The mainly problem is the algorithm is single. It can make better use of the advantages of different algorithms by combining different optimization algorithm to design the system, and then the system performance is better and more suitable for the application. It is a new topic of the present study. The thesis starts from the research based on design algorithm of three kinds of interval type 2 TSK fuzzy logic systems including A1-C1, A2-C0 and A2-C1(where A denotes antecedent parameters, and C denotes consequent parameters), and designs three type-2 TSK mixed system models. Moreover, the designed hybrid models are applied to forecast practical problem, and the simulation study is given. The simulation results show that the designed hybrid models are feasible and effective for practical problems. The specific work is as follows:(1) Introduce detailed the fuzzy logic systems, TSK fuzzy logic systems and the related parameters optimization algorithm.(2) Study the hybrid learning algorithm based on A1-C1 interval type-2 TSK fuzzy logic systems. The fuzzy logic systems are integrated into neural network to form the fuzzy neural network systems. Fuzzy k-means clustering method is applied to filtering rules. Firstly, apply least square method to adjust the system consequent parameters. Secondly, apply BP algorithm to adjust the system antecedent parameters. Finally, the designed hybrid system models are applied to forecast the price of International Brent oil, and the simulations are performed with MATLAB. Then make a comparison between the designed hybrid algorithms and a single BP algorithm. The simulation results show that the designed hybrid models are applied feasible and effective to forecast the practical problem.(3) Study the hybrid learning algorithms based on A2-C0 interval type-2 TSK fuzzy logic systems. The fuzzy logic systems are integrated into neural network to form the fuzzy neural network systems. Fuzzy k-means clustering method is applied to filtering rules. Firstly, apply least square method to adjust the system consequent parameters. Secondly, apply BP algorithm to adjust the system antecedent parameters. Finally, the designed hybrid system models are applied to forecast the price of International Brent oil, and the simulations are performed with MATLAB. Then make a comparison between the designed hybrid algorithms and a single BP algorithm. The simulation results show that the designed hybrid models are applied feasible and effective to forecast the practical problem.(4) On the basis of A1-C1 and A2-C0 interval type-2 TSK fuzzy logic systems, study the hybrid learning algorithms based on A2-C1 interval type-2 TSK fuzzy logic systems. The fuzzy logic systems are integrated into neural network to form the fuzzy neural network systems. Fuzzy k-means clustering method is applied to filtering rules. Firstly, apply least square method to adjust the system consequent parameters. Secondly, apply BP algorithm to adjust the system antecedent parameters. Finally, the designed hybrid system models are applied to forecast the price of International Brent oil and Russian Trading System Index, and the simulations are performed with MATLAB. Then make a comparison between the designed hybrid algorithms and a single BP algorithm. The simulation results show that the designed hybrid models are applied feasible and effective to forecast the practical problem.