A Class Of Time-varying Weight Interval-valued Time Series Combination Forecasting Method And Its Application

As one of the most basic energy and chemical raw materials suppliers,crude oil plays an important role in the economic and social development and the improvement of people's livelihood.Crude oil has been the focus of attention around the world for years.In recent years,the market price of international crude oil fluctuates more violently.In order to better grasp and study the useful information contained in the international crude oil price and the uncertainty actually contained in the data,this paper chooses the interval-type international crude oil price data as the research object for analysis.From the perspective of interval valued time series,this paper combined with interval valued time-varying weight combination prediction model to carry out the prediction research.According to whether the weight of portfolio prediction changes with time,the portfolio prediction can be divided into the form of fixed weight and time-varying weight.Firstly,in view of the problem that there are few researches on the combination prediction of interval value time-varying weight,this paper proposes a construction model of interval value time-varying weight based on the construction method of real value time-varying weight.Secondly,in order to further improve the prediction accuracy,a sliding window mechanism was introduced to construct a time-varying weight model.Then,the traditional sliding window construction time-varying weight method has the problem of raw data loss to a certain extent.In this paper,an improved sliding window time-varying weight prediction model is proposed and combined with interval time-varying weight results,a class of optimized time-varying weight interval value sequence combination prediction model based on sliding window is obtained.Finally,in view of the problem that most of the current combination prediction results do not involve or realize trend extrapolation,this paper will establish the component data prediction model combined with the idea of component data,so as to realize the extrapolation of the final interval value combination prediction results.A kind of time-varying weighted interval value time series combination forecasting method proposed above is empirically analyzed by using international crude oil price interval data.The results show that: 1.The validity of the prediction results of the interval time-varying weight combination is better than that of the single prediction results.2.Compared with the traditional method,the prediction accuracy of the improved sliding window time-varying weight prediction model is improved and the loss of original data is avoided.3.After combining the improved sliding window time-varying weights with the results of interval value time-varying weights,the accuracy of a class of optimized combined prediction models is improved compared with that before the combination,in which the interval optimization weighting method and the improved sliding window weighting method get the best result of the time-varying weight combined prediction model.4.The extrapolation prediction model of component data is constructed according to the optimal time-varying weight sequence,and compared with the final extrapolation results,it is found that the prediction accuracy of the extrapolation model is significantly higher than that of single prediction results.In this paper,the innovation points mainly include the following: 1.This paper extends the determination method of time-varying weight combination prediction weight to the interval value sequence,and then establishes the interval value time-varying weight model to realize the construction of the interval value time-varying weight combination prediction model.2.The sliding window mechanism is introduced from combination prediction and the time-varying weight is constructed according to the sliding window principle.On the basis of this,an improved time-varying weight model with interval values of sliding window is proposed.3.In order to further improve the prediction accuracy and make up for the deficiency of time-varying weights of interval values under different methods,the time-varying weights under different methods are weighted and averaged.An optimized time-varying weighted interval-valued combined prediction model is obtained.