A theoretical study and forecasts of fuzzy time series
Chao-Chih Tsai · Journal of Interdisciplinary Mathematics · 2001
The theory for analyzing the model equation is proposed in this study. It clarifies the algorithm suggested by Mamdani [6] used to calculate the fuzzy relational matrix and indicates the error of the matrix may be caused. Thus, we can find the mechanism for improving. A new formulation is then presented for the modeling of high-order fuzzy time series. To illustrate the current model, the forecasting populations are carried out. the data of historical populations in Taiwan area are adopted. The results are compared with those of the traditional regression method and those of Song-Chissom method [5]. It is found that the root mean square error of the forecast can be improved from 0.119 for the Song-Chissom method to 0.013 for the second-order model. The proposed model is more accurate, but retains the simplicity and robustness.