Applied Research Of Non-linear Method Time Series Combination Model In Agricultural Product Price Forecasting

As an ancient country with a long history of agricultural civilization,the issue of rural areas,agriculture,and peasants has always been a major issue in the country's work.It is the duty of all Party workers to properly resolve the issue of agriculture,rural areas and farmers.As an important part of the agricultural product market in the modern economic system,the price of agricultural products is closely related to the vital interests of farmers and even the quality of life of the people.It is also one of the important livelihood issues.Therefore,the forecast of the price of agricultural products has become the focus of research in recent years.As the price of agricultural products is influenced by external factors such as production costs,changes in supply and demand,natural climate,and government policy adjustments,the frequency of price fluctuations is rapid and large,showing a non-linear and non-stationary character,which brings about the production of farm households.Great risk.In order to meet the urgent needs of agricultural product producers,operators and government-related decision-making management departments,many scholars have conducted exploratory research and case validation on the prediction of agricultural product prices in recent years.However,regardless of whether single prediction method or combination forecasting method is used,there are two major defects in these researches.First,data acquisition is difficult,sample input requires production cost,supply and demand changes,natural climate,government policies,and other multidimensional time series.However,these related data are difficult to obtain in a full range of operations in practice.Second,the process of selecting a model to predict a non-linear,non-stationary raw time series directly ignores the filtered random terms.I think this is also One of the reasons for the poor prediction accuracy.After comprehensively understanding the research status of agricultural product price forecast at home and abroad,the current time series combination model cannot predict the random items of the sequence well in nonlinear time series forecasting and it is difficult to obtain data.For the problem,a time series autoregressive combination forecasting model based on nonlinear methods is proposed.Firstly,the wavelet analysis method is used to decompose the time series of the original agricultural product price to obtain several layers of high-frequency components and low-frequency components.However,as the decomposition scale increases,the corresponding number of data decreases rapidly,so the low-frequency decomposition vector and high frequency are decomposed again.The decomposition vectors are respectively reconstructed into two independent high frequency sequences and low frequency sequences.Secondly,two time series autoregressive prediction models based on non-linear methods are used to predict the two time series.Non-linear non-recurrent neural network(NARX)prediction models are used to predict low-frequency sequences with slow fluctuation amplitudes.Generalized self The regression condition heteroskedasticity prediction model(GARCH)is used to predict high-frequency sequences with large fluctuation amplitudes.Finally,these two parts of the prediction results are linearly synthesized to obtain the final predicted value.This combined forecasting method not only ensures the comprehensiveness of information,but also makes full use of the autocorrelation of agricultural product price time series.The experimental results show that the GARCH-NARX combined prediction model has better adaptability and higher prediction accuracy for a single nonlinear non-stationary time series,which verifies the feasibility of the method and provides a new prediction in theory.