Research On Time Series Forecasting Based On Spatial-temporal Fuzzy Information Granules

Time series refers to continuous observation series based on time dimension,which contains rich information of the observed system.The research of time series prediction is to obtain the inherent evolution of data from the time series of prediction index by using statistical method or intelligent technology,and to make quantitative estimation for the future development and changes of forecasting indicators,which are generally used in many fields such as engineering and finance.Because the observation data obtained in the real world are generally fuzzy and other uncertain features,it has important application value to introduce the fuzzy set theory into the field of time series prediction and establish the fuzzy time series prediction model.Therefore,based on the existing research work,this paper puts forward the construction method of constructing spatial-temporal fuzzy information granule,analyzes and studies the long-term prediction of time series,mainly including the following aspects:1)Spatial-temporal fuzzy information granule(STFIG)is proposed to mine characteristics of data from two dimensions of time and space,so as to improve the interpretability of information granule.In the time dimension,the variable length segmentation method of stepwise linear division(SLD)is used to describe the trend information and dispersion degree of sequence data.In the spatial dimension,the change rate of time series is calculated to represent the spatial fluctuation information.At the granularity level,time series prediction is extended from single step to multi-step,so as to achieve the goal of long-term prediction of time series.2)Compared with the type I fuzzy system,the type ? fuzzy system can describe multiple uncertainty information.For the time series generated by the complex nonlinear system with high noise,in order to further improve the processing ability of its fuzzy uncertainty and noise,the interval type-2 fuzzy set(IT2FS)is introduced into the granularity model.On the basis of ensuring low computational complexity,the expression ability of fuzzy information is enhanced,and the validity of long-term prediction model is improved.3)Based on the Hausdorff distance theory,a distance algorithm is proposed to measure the similarity between two spatial-temporal fuzzy information granules,and multiple groups of different real experimental data sets are used to verify the effectiveness of the algorithm,so as to improve the stability and accuracy of the forecasting effect.