Research On Pork Price Prediction Based On Feature Selection And LSTM Hybrid Model

At present,there are many factors that affect pork prices in the context of the epidemic,and there is a non-linear relationship between wholesale prices and fluctuations,which poses new challenges to the prediction of pork prices.In order to accurately and effectively predict the wholesale price of pork,this paper proposes a PCA(principal component analysis)-CNN(convolutional neural network)-LSTM(long-term and short-term memory network)prediction model,which can well present the cyclical changes of pork prices,effectively predict the price trend and its trend,provide price reference for pork sellers,and enable the government to better regulate and supervise the pork market.This paper selects the daily pork price from February 2022 to February 2023 as the sample data,uses RF(random forest)to extract the characteristics of eight characteristic attributes,including pig(internal ternary),pig(local hybrid pig),pig(external ternary),corn,soybean meal,beef,white mutton,chicken,and then uses PCA(principal component analysis)to realize the dimension reduction and feature optimization of factors affecting the wholesale price of pork,Finally,one principal component with a cumulative contribution rate of over 88% to the wholesale price of pork is input into the CNN-LSTM network model to effectively predict the wholesale price of pork.Experiments have shown that compared to traditional models and shallow prediction models that have not been extracted and optimized for features,the PCA-CNN-LSTM neural network model has higher prediction accuracy.