Research on Long Short-Term Decision-Making System for Excavator Market Demand Forecasting Based on Improved Support Vector Machine

Abstract

Future demand forecasting of the excavators is of great significance to guide the supply and marketing plan. For a long time, market forecasting of the construction machinery is regarded as short-term forecasting, which lacks the analysis of macro-marketing law and cannot reflect the true law of market development. In this paper, a decision-making system based on both long-term and short-term features was proposed. The interval classification and recursive feature elimination were used to select the main factors that affect the demand of excavators. Then a support vector regression model based on decomposition synthesis (DS-SVR) was developed to forecast the long-term features, and a model combined with a seasonal autoregressive integrated moving average model (SARIMA) was developed to forecast the short-term features. Finally, the differential evolution algorithm (DE) was applied to optimize model parameters. The performance of the forecasting model was tested using the marketing data of a typical enterprise. The results showed that the total error rate of the forecasting model for the one-year long-term forecasting is 26.61%, and the classification error of forecasting of the three-month short-term forecasting are 13.65%, 18.83%, and 19.62%, respectively, which are superior to the SVR forecasting model and the SARIMA forecasting model.