The Time Series Prediction And Application Research Based On IT2 TSK FNN System

Since information is explicit in the fuzzy logic system, the fuzzy logic system is simple, its algorithm is direct and its processing capacity is small. But it doesn’t have the ability to learn by itself and its agility is very poor. The insufficient part of neural network is mainly displayed in the following three aspects: difficult understanding, complex structure and large amounts of information. However, its advantages are listed below: the ability of learning by itself, a high level of parallelism and excellent robustness. In this thesis, the fusion between the fuzzy logic system and neural network will surely break the limitations existing in the single system. This thesis puts the Type-2 TSK fuzzy logic system and neural network together to form a kind of Interval Type-2 TSK fuzzy neural network system. The system’s function is similar to neural network and Type-2 TSK fuzzy logic system. This thesis chooses BP algorithm(Error Back Prorogation algorithm) to solve the problem of the Shanghai Composite Index prediction successfully using the designed Interval Type-2 TSK fuzzy neural network system. Specific works are as follows:Research the Type-1 TSK fuzzy neural network system based on Type-1 TSK fuzzy logic system and neural network, select BP algorithm to adjust the various parameters of the model, combining with the Shanghai Composite index prediction application examples, and use MATLAB program to do stimulation tracking. The result shows that the specially designed Type-1 TSK fuzzy neural network is effective and feasible.After researching the interval Type-2 TSK fuzzy neural network system based on Type-2 TSK fuzzy logic system and neural network, we successfully designed the structure of the interval Type-2 TSK fuzzy neural network system. We still choose BP algorithm to adjust the various parameters of the system model, predict the Shanghai Composite index using the designed Type-2 TSK fuzzy neural network system, and use the MATLAB to simulate the effect of tracking. The result shows that the Type-2 TSK fuzzy neural network system designed in the thesis is reasonable, feasible and effective. The figure of stimulation effect of tracking and the root mean square deviation show, that the Interval Type-2 TSK fuzzy neural network system has better controlling performance than Type-1 TSK fuzzy neural network system when there exists uncertainty in practical issues.