Forecast of Tobacco Raw Material Demand Based on Combination Prediction Model

Abstract

In order to improve the prediction accuracy of tobacco raw material demand, this paper presented a combined prediction model. Combined prediction model first used Holt-winters exponential smoothing method and SARIMA model to forecast the demand of cigarette raw materials respectively, and then used BP neural network to aggregate the results of these two predictions to get the final prediction result. Holt-winters exponential smoothing method, SARIMA model and combined prediction model were used to forecast the demand data of tobacco raw materials, respectively. For the prediction of the same material, the error of the combined prediction model were all less than the other two models. The prediction accuracy of combined prediction model was higher.