Hits:

Indexed by:会议论文

Date of Publication:2014-01-01

Included Journals:CPCI-SSH

Page Number:29-35

Abstract:A computational method of forecasting based on BP neural network and high-order fuzzy time series is presented in the paper, which provides a flexible method for recognizing the fuzzy relation R of the time series. The proposed method is a high-order model and as an example the computation is implemented with order 3. The proposed model has been tested on the historical student enrollments of University of Alabama to compare with the existing methods. The better forecasting results are obtained with the average forecasting error 1.129% and mean square error 51916, which are smaller than the ones presented in [2,3,4]. It is a pity that the forecasting results are not always the same due to the randomness of the forecasting system so the choice of the initial weights is important.