Research On Fuzzy Time Series Model And Application In Distribution Of Mobile Business

The mobile business volume is growing fast along with the constant innovation and development of communication technology, while the mobile resource is limited. Accurate forecasting for the volume of mobile business can offer basis for effective allocation of wireless resource and assure the great quality of service for subscriber. So forecasting the volume of mobile business is significant.Time series analysis is a popular method of numerical prediction.While traditional time series models get bad performance without the situation of uniform noise and smooth data. Fuzzy time series not only can achieve better performance under the circumstance of fluctuant noise and non-constant relationships between several adjacent data, but also is simple to apply, which is why it is popular for being applied in network business volume.In this paper, the research on fuzzy time series model is done firstly.A fuzzy time series forecasting models based on fuzzy logical relationships tree are proposed, in which fuzzy logical relationships tree applies first order, second order, third order fuzzy logical relationships into forecasting, it utilizes the robust forecasting information of fuzzy logical relationships of low orders and precise forecasting information of fuzzy logical relationships of high orders to improve the accuracy of forecasting. Considering the advantage of application of relevant information between several series to assist forecasting, fuzzy logical relationships tree is exploited into multi-variable fuzzy time series model,in which a novel algorithm for dividing the universe of discourse and heuristic rules are proposed. The experimental result shows that the second model gets better forecasting performance.Moreover, the proposed fuzzy time series model is exploited to forecast the mobile business volume based on above research. The time series are constructed by the certain time granularity selected from the measure report of ten cells in hot network area. According to the experimental results, the proposed model gets great forecasting performance in the situation of mobile business.