Fuzzy time series forecasting based on information granule and neural network

Abstract

Time series forecasting is critical for the research of a fuzzy time series. In this paper, a novel model combined information granule partitioning method with back propagation neural network (BPNN) is proposed to forecast the time series. First, the unequal-dividing method based on information granule is applied to divide the universe of discourse to form the fuzzy sets. Then, we use the fuzzy sets to fuzzify the historical data into labels. Next, a second-order fuzzy logical relationship of the labelled dataset is constructed to train a BPNN to forecast the labels. Finally, the forecasting labels are defuzzified to obtain predictions. The Taiwan Stock Exchange Capitalisation Weighted Stock Index (TAIEX) is used to verify the effectiveness of the proposed model. The results show that the proposed model performs better than existing models according to root mean-square error (RMSE).