The Research Of Multi-factors Fuzzy Time Series Forecasting Models

The study of fuzzy time series has a great advantage of its salient capabilities of tackling uncertainty and vagueness inherent in data collected, So that more researchers are attracted to build fuzzy time series forecasting model to solve the problem. Fuzzy time series models are widely used in various fields and high prediction accuracy has become the focus of attention of researchers. The traditional fuzzy time series models mostly handle the model of single-factor problem, however, there is nearly no phenomenon is independent; it has relationships with other factors more or less. How to build multi-factor fuzzy time series model has become an urgent problem. The objective of this paper is to study this problem. In the fuzzy time series models, the establishment and application of universe of discourse partitioning, data fuzziness and fuzzy relationships are the three important parts.In the universe of discourse partition. In the classic fuzzy time series model usually employ universe of discourse to constructed fuzzy sets, Appropriate intervals and their lengths are always challenging problems and researchers attempt to improve accuracy of fuzzy time series by selecting proper intervals and adjustment of their lengths. Based on above, The paper presents a theory of space FCM to partition the universe of discourse, it reflects better about the multi-factor date in space distribution by space distance. And it also reflects the true date structure and the sensitivity of the data. Finally, it is forecasted by the Amazon stocks close price, the high price and the low price, and it verifies the effectiveness of the proposed algorithm.In the fuzzy relationship establish and apply. Based on the analysis of classic fuzzy time series forecasting models, this paper proposes based on partitioning theory and evidence theory of multi-factors fuzzy time series model, and based on spatial FCM and multi-factors fuzzy time series model of evidence theory, evidence theory based on the adjustable parameter of multi-factors fuzzy time series model. And finally verify the feasibility and effectiveness of the proposed theoretical model by the Amazon stock close price forecasting result in this paper.