Application of Machine Learning Methods in Pork Price Forecast

With the improvement of people's living standards, people's consumption of meat is getting higher and higher, and pork has become the core of Chinese meat production and consumption structure. Among pig farmers, retail investors account for more than half, their risk resistance capacity is weak, and they are vulnerable to price shocks. The price of live pigs showed significant seasonal changes, and violent fluctuations not only affected the interests of various links in the pig industry chain and the welfare of consumers, but also affected the development of the entire Chinese pig industry. Effective hog price forecast which is conducive to social stability and unity can not only ensure the income of farmers, but also ensure relationship between supply and demand. The article synthesizes the main indicators related to pork prices in the Chinese pork market, applying DBN (Dynamic Bayesian network) method and the SVM (support vector machine) method, the BP neural network method, these Machine Learning methods, and compare with traditional methods of the ARIMA method, to establish a predictive model of pork prices. The experiment was conducted in R and Bayes Server using 2001-2016 price data from the National Bureau of Statistics. The price is forecasted and analysed, the prediction effects of the four models are compared in this paper. The results show that the accuracy of predicting the pork price based on DBN model is better than other methods, RMSE=1.200822, MAPE=1.137312, TIC=0.0351875, all belong to a minimum.