The Study On Grain Price Prediction Based On Lasso And Support Vector Machine

Food as a product of key projects,its price fluctuations will influence food production,food related products import and export of our country as well as the national macroeconomic regulation and control,which has a far-reaching influence to consumers and producers.Therefore,understanding and analyzing the various factors that affect the food price volatility,the factors affecting the related regression prediction model is established on the basis of the analysis and prediction of grain price change trend,on the one hand,can understand the change of China's grain price trend;On the other hand can provide relevant price regulation and macroeconomic policy to provide scientific reference for the stability of grain prices and ensure food security has important practical significance.This paper according to the theory of fluctuations in the price of food and the comb before scholars on the study of factors affecting food prices,21 variables were selected to measure the influence factors of China's grain price fluctuations.Then the Lasso method and support vector machine(SVM)the characteristics and application range of the two methods are analyzed and summarized,we found that the two methods into use in the food price projections have their own advantages and disadvantages.So in this paper,the Lasso method and support vector machine to complete the fitting and forecasting of grain prices,make up for their respective shortcomings in the prediction process,to achieve their complementary advantages,make the prediction results to achieve better accuracy,mainly gives the possibility and the way in the following three fusion.First,the Lasso method combined with support vector machine(SVM)by tandem manner,its core idea is:first use Lasso method selected the main factors affecting the food price volatility,and the main influencing factors as input of support vector machine(SVM)model,through constant learning and training,it is concluded that the grain price forecast.Second,the Lasso method combined with support vector machine(SVM)through a parallel manner,namely respectively using Lasso method and support vector machine(SVM)to predict,and then under the condition of minimum mean square error to find the optimal combination weights will combine two forecast results.Third,the Lasso method are combined with support vector machine(SVM)by embedding method,its basic idea is to Lasso model,support vector machine(SVM)model of grain price forecasts as input vector of SVM prediction model,the corresponding food prices,the actual value of the moment as the output target,establish a portfolio model to predict sample to finally get the food price forecast.Finally,the combination model and the method of single model applied to food price forecasting practice,through the comparison of predictive value,found that based on the Lasso method and support vector machine series combination model and the embedded type combination model prediction accuracy of the forecast accuracy was significantly higher than the other four,proves that using the Lasso method and support vector machine(SVM)to food prices for the feasibility and advantage of combination forecast.