Skip to Main Content
The principles of support vector regression (SVR) are described. The collection and treatment of customer demand, the moving time window method, the selection of training samples and the analysis of forecasting accuracy are stated. The customer demand forecasting approach based on SVR using moving time window method is proposed. With the demand data of a simulation example, the presented approach is used to forecast the demand values for 7 days ahead. The average forecasting error is less than 2%. The simulation results demonstrate the approach is feasible and valid in customer demand forecasting.