Research On Agricultural Product Price Forecast Based On Combination Model

In the modern economic system,the agricultural product market is an indispensable and important component,and the price of agricultural products is directly related to the vital interests of farmers and closely related to people’s lives.This article first explores the prediction methods of traditional agricultural product prices,and proposes an improved method based on thorough analysis and research of traditional prediction methods.Traditional agricultural product price forecasting only uses a single model,such as statistical methods based on time series,autoregression or machine learning.In recent years,the method of using machine learning combination model has been more and more studied.However,most agricultural product price data has characteristics such as complexity and nonlinearity,which limits the accuracy and model fitting ability of traditional agricultural product price prediction methods.This article takes the price of live pigs in China as an example and proposes a combination model for predicting agricultural product prices.Firstly,robust local weighted regression(STL)is used as a smoothing method to decompose the price time series;Secondly,the decomposed trend and residual component data are trained with the Long short-term memory Memory(LSTM)network to obtain two sets of predictive values.The seasonal component data is trained with the Seasonal Autoregressive Integrated Moving Average(SARIMA)model to obtain another set of predictive values.The three sets of predictive values are added to obtain the predictive values based on the STL time series decomposition;Then,the Attention Gate Recurrent Unit(Att GRU)model is used to predict pig price data and influencing factor data to obtain predictive values based on multivariate influencing factors;Finally,the predicted values obtained from the STL time series decomposition and the Att GRU model are fitted using a Back Propagation(BP)neural network to obtain the final predicted values.This combination prediction method not only fully utilizes the autocorrelation of agricultural product price time series data,but also ensures the integrity of the impact information of price correlation.Comparing the prediction results of the STL-Att GRU-BP combination model with the other six models,it was found that the STL-Att GRU-BP model produced the lowest RMSE,MAE,and MAPE,and the highest R~2,achieving more accurate and stable predictions of agricultural product prices.At the same time,the method mentioned in this paper has the potential to be applied to price forecasting of other agricultural products,which will help the development of precision agriculture and digital agriculture.