Prediction Of Hog Price And Produce Based On Dynamic Bayesian Network

China is the world's largest producer and consumer of pork,and the world's largest producer of pigs.China's annual pig production is 600 million.In all meat consumption in China,pork consumption has been ahead of other meat products.In 2016,pork consumption accounted for 60% of meat consumption.In recent years,hog prices have not only fluctuated frequently but also have increased volatility.The fluctuation of hog prices will have a huge impact on the entire pig industry and society.It can be seen that studying a model that can accurately predict the hog price is great significance for the research and production of the pig market.The results of hog price forecasting can not only provide decision support for the government to adjust all aspects of the pig industry chain,but also balance the food consumption price of urban residents,ensure the balance of supply and demand in the pig market,and promote the stable development of China's pig industry.The factors affecting the hog price are complex and uncertain.This paper firstly uses the correlation analysis to select the important influencing factors from the 20 kinds of influences of hog price fluctuations as the input parameters of the hog price forecasting model.Then,based on the static Bayesian network,the time series is combined,and the model is established through structure learning and parameter learning.This model is a dynamic Bayesian network model suitable for solving uncertainty problems.The specific process is as follows:1)Data processing.In this paper,Python web crawlers are used to obtain experimental sample data.After sample data is collected,the data is preprocessed,including missing values and culling outliers.According to the experimental requirements,the sample data of different dimensions are normalized,and the data is divided into training set and test set and used in the experimental model.2)Variable selection.The influencing factors of the price of hog are complex and diverse.According to the reference literature and expert knowledge,more than 20 factors are divided into important degrees,and the factors affecting the hog price are analyzed.The important influencing factors are selected,namely the price of soybean meal and the price of piglets,pork production,pig slaughter,pig stocks,ex-factory price of pigs,pig-to-food ratio,consumer price index of livestock and meat,wholesale price of white pigs,wholesale and retail profits.3)Build the model.The construction of dynamic Bayesian network model is divided into structure learning and parameter learning.The structure learning of this paper adopts three methods,including learning from data using PC algorithm,constructing according to expert knowledge and combining expert knowledge with PC algorithm.After learning,the initial structure is obtained and adjusted to obtain the network structure of the model.Then the EM algorithm is used to learn the parameters,and a complete dynamic Bayesian network model is obtained for price prediction.It is predicted that the effect will be the best model.In order to verify the prediction effect of dynamic Bayesian network model,this paper also establishes a prediction model based on ARIMA,BP neural network and Support Vector Machine(SVM)to predict the price of hog,using Root Mean Square Error(RMSE)and Mean Absolute Percentage Errors(MAPE)and Hill Inequality Coefficient(TIC)are used as evaluation criteria to evaluate the prediction results of the above four models.The results show that the DBN model has an RMSE=1.200822,MAPE=7.137312,TIC=0.0351875 in the hog price forecast,both are superior to other models.The DBN model predicts that the hog price in 2019 and 2020 is 13.85557 yuan in 2019 and 13.6012 yuan in 2020.In the production forecast,the DBN model has RMSE=37.611352,MAPE=0.626567,TIC=0.003349 and is excellent.In other models,the DBN model predicts that the hog production in 2019 and 2020 will be 546.864 million tons in 2019 and 548.219 million tons in 2020.