A New TSK Gaussian Fuzzy Inference System with Principal Component Weight

Nowadays, fuzzy inference systems provide extremely effective support in solving many problems in human life. Among them is the strong performance and application of Takagi-Sugeno-Kang (TSK) fuzzy inference system. In this study, a new Gaussian fuzzy TSK system using principal component analysis method is proposed to reduce the capacity of fuzzy rule system when the number of input criteria is relatively large, and at the same time, the model has improvements when considering the weight value of input criteria by determining the contribution proportion of input information. The model uses the entropy minimization approach (MEPA) to provide effective support for the fuzzification process of series input data. The proposed model is applied to forecast the socio-economic development index of 63 provinces in Vietnam, and the input data are the socio-economic development indicators of 63 provinces in 2019. The forecasting results of the proposed system are evaluated and analyzed by comparing the forecasted and actual values, and evaluating the values of MSE, RMSE, MAPE and CORR between the model using weighted and the model not using weighted values. Using weighted values shows that the proposed model has much better performance.