The Application Of Fuzzy Time Series In The Automotive Sales

In recent years, many scholars at home and abroad dedicated to the relevant research of predicting data, for example stock, futures and other fields have been harvested. Partitioning the universe of discourse and determining effective intervals are critical for forecasting in fuzzy time series. Equal length intervals were proposed and used in most existing literatures. If equal length intervals are used for the data that has good linear or strongly regularity, the forecast will be more accurate. But they are used for the data that has strongly fluctuation and nonlinear or irregular data, such as stock, futures and so on, the precision of the prediction will be discounted. In this background, the proposed method raises in this paper.The focus of this paper is that unequal length intervals division method is put forward, to improve the accuracy of prediction, which unequal length intervals division method is implemented by the theory of information granule. First, we calculate the prototypes of data using fuzzy clustering, then through using information granule to adjust interval. The last, we make forecast according to the established seven interval and relevant theories of fuzzy time series.In this paper, we study how to partition the universe of discourse into intervals with unequal length to improve forecasting quality. First, we calculate the prototypes of data using fuzzy clustering, then according to the prototypes to form several subintervals. These intervals not only need to partition reasonably according to the data partition, but also need to satisfy the rationality of semantics. In order to prove the proposed method correctness and effectiveness, this paper uses the car company sales data for simulation and experiment. Finally, we use MSE for evaluating the accuracy of prediction. Further more, the proposed method is very robust and stable for forecasting in fuzzy time series. In the future, we should apply the method to other fields, and also add some variables. If that, forecasting will become more diversity.