Research On A Method Of Forecasting Grain Price Based On Dynamic Exponential Smoothing Model

The fluctuation of grain price and the future trends are closely related to people’s dailylives. In-depth analysis and mining large numbers of transaction data and making priceforecast are conducive to managing and directing grain trade, guiding agricultural producersto choose the right crops according to the specific conditions, promoting the informatizationand marketization of agricultural production, reducing the costs and economic risks of cropsproduction. Therefore, it’s of vital significance to study and forecast price data of agriculturalproducts. In recent years, as an important branch of time series forecasting analysis,exponential smoothing’s features of superior performance, applicability and using easily makeitself have been developed rapidly and researched in depth in the field of forecasting. And ithas been used widely in military, science, economy and so on.With in-depth research and extensive application, the researchers found that there arethree problems to solve in the exponential smoothing model. First of all, the smooth initialvalue is difficult to determine. Follows, it’s difficult for the static smoothing parameter toadapt to the change of time series. Finally, the value of smoothing parameter is mostly fromresearchers’ experience and many experiments, so it is often difficult to obtain the best values.The research emphasis in this thesis is to solve the difficulties of traditional exponentialsmoothing model, and apply it to the field of grain price forecasting.Firstly, this thesis analyzes the theory of exponential smoothing model, and discusses theadvantages and disadvantages of the original model. Secondly, on the basis of previousresearch results, through in-depth analysis and complete demonstration, it was improvedbased on original quadratic smoothing formula model. The establishment of dynamicexponential smoothing model based on dynamic parameters overcomes the shortcomings ofstatic smoothing parameter’s difficulty in adapting to the time series changes in forecastingprocess. And the introduction genetic algorithm to exponential smoothing algorithm optimizethe selection of parameters, solving the difficult problem of smoothing parameter determining,providing better support for decision-making. Then, this thesis uses truthful data of grainmarket to do contrast experiment on double exponential smoothing model, Holt model and thedynamic exponential smoothing model. The comparison of experimental results confirmsthe superiority of the dynamic exponential smoothing model and the prediction accuracy ratehas more significant improvement than the double exponential smoothing model. It hasrobustness in the three forecasting models. Finally, this thesis summarizes the results of this research, and looks forward to future research directions and priorities in grain priceforecasting and the application in other areas.