Prediction Analysis Of Pork Price Based On GA-BP Neural Network Model

With the continuous progress of reform and opening up,China’s agriculture,especially pig breeding industry,continues to move to a new stage.With the continuous acceleration of pig scale process,pig breeding has become the pillar industry of China’s agriculture.The number of pigs and pork production in China have increased year by year,and the total annual pork consumption can reach half of the global annual pork consumption.However,in recent years,due to the influence of macroeconomic,national policy adjustment,epidemic disease and many other factors,China’s pig breeding industry is facing great challenges.The supply of pigs is sometimes sufficient and sometimes short.Such a repeated phenomenon has brought troubles to both consumers and farmers.Therefore,accurate prediction of pork price has very important positive significance.Through sorting out the relevant researches of domestic and foreign scholars on pork price,it is found that most foreign scholars start from the law of pork price fluctuation and seldom make specific prediction and analysis on pork price.However,domestic scholars mainly use various forecasting models to make empirical prediction of pork price.The main work of this paper is as follows:1)Deeply analyze the rules of pork price fluctuations in Recent years in China and summarize its internal mechanism,fully draw on the advantages of relevant research of domestic and foreign scholars,and summarize the main factors affecting pork price fluctuations through the analysis of pig production,market supply and demand relationship and market external factors,including 14 indicators.They are piglet price,pig price,pork output,pig stock quantity,market quantity,chicken price,egg price,beef price,mutton price,corn price,soybean meal price,fattening pig feed price,pig epidemic rate and household disposable income.2)The data were preprocessed,and the national average pork price from the 1st week of 2015 to the 52nd week of 2021 was taken as the target variable.NPCA method was used to reduce the dimension of 14 influencing factors of pork price,and finally two principal components were extracted for subsequent modeling prediction.3)In order to compare the prediction performance of GA-BP model after NPCA dimensionality reduction,this paper constructed another four comparison models,namely GA-BP model after PCA dimensionality reduction,GA-BP model without dimensionality reduction,ARIMA model and SVR model,and substituted the sample data into the corresponding models for fitting prediction.4)Comparative analysis of the predicted results.Draw the curves of the predicted results and expected results of each model,calculate the error of the predicted results of each model,and deeply analyze the prediction performance of each model.Finally,it is concluded that NPCA-GA-BP model has the best effect on pork price prediction,and this model can be used to predict the future trend of pork price.The innovation of this paper is as follows:1)Chinese scholars mostly use PCA method to reduce the dimension of the influencing factors of pork price.In this paper,after comprehensively considering the characteristics of pork price fluctuation,NPCA method is used to reduce the dimension of the original sample data,which achieves a better effect of dimension reduction.2)The analysis shows that pork price prediction is a nonlinear problem.In this paper,the superposition of nonlinear principal component analysis and BP neural network is used to better analyze the nonlinear problem.Shortcomings of this paper:In the data selection,only some available data of influencing factors of pork price are selected,but in fact,the pork market is very complicated,such as the amount of pork reserve released by the state,pork import and export volume,international pork price,macroeconomic policies and other factors will have different degrees of influence on the pork price in China.However,the data of some factors are difficult to obtain or some factors are difficult to quantify,so they are not taken into account in the prediction research of this paper.