Pork Price Prediction Modeling Based On Feature Selection And LSTM

Pork has the basic attributes of people’s livelihood in the country,and it is also one of the most important sources of meat for urban and rural residents in China.In recent years,its price has been affected by many factors such as internal and external environment,and the price has fluctuated,output and supply have fluctuated violently.At the same time,the African swine fever epidemic has swept the world,and the new crown pneumonia has affected the healthy development of China’s domestic pig industry.It has had a great impact on the daily and economic life of the people of the country.In order to be tter explore the prediction method of pork price,through the analysis and screening of 14 characteristic variables that affect pork price,a mathematical model with Long Short-Term Memory Network(LSTM)as the core is established with the aid of computer.Model to make scientific predictions of pork prices.Discuss based on decision tree algorithm(Decision Tree,DT),adaptive boosting algorithm(Adapyive Boosting,Adaboost)That merger pattern about RF,GBDT and LSTM,the main research contents of this paper are as follows:1)Collected and established 14 indicators and 9,315 basic data affecting pork prices,and carried out qualitative comparison and analysis of pork prices in the same period for the 14 basic data.Next,four types of machine learning algorithms are used to calculate the feature importance that affects pork prices,and a series of reference conclusions are drawn.With the feature importance ranking of the target variable(pork price),the research will leap from qualitative analysis to quantitative analysis has important practical significance to farmers,consumers and relevant government departments.2)LSTM was introduced to predict pork prices,The training set was 60% of the data,the testing set was 40% of the data to predictive pattern MAPE、 MAE、MSE、MSE are used as measurement indicators.So that comparing the LSTM’s prediction consequence,several common algorithms are selected for prediction resolvation and constration.Consequence it is shown that compared with other machine learning models,LSTM has a better prediction effect,so that further promote the vaticination preciseness of the pattern.3)When we begin to choose variables listed in the text which have a significant impact on pork prices are selected as much as possible,and the combination of decision tree and LSTM,the combination of random forest and LSTM,the combination of Adaboost and LSTM,GBDT,LSTM combination,through experiments,this paper can get which combination of features can get the best pork price prediction.When we use the LSTM pattern to contrast the performance of the LSTM pattern without screening variables,it can be intuitively seen that the prediction model after feature screening has higher pred iction accuracy.