Combination Prediction Model for Logistics Demand Based on Least Square Support Vector Machine

摘要

Combination prediction is an effective method to improve prediction precision for logistics demand. On the basis of least square support vector machine (LS-SVM), a combination prediction model for logistics demand is proposed. Firstly, according to the historical data of logistics demand, grey model (GM), auto regression moving average (ARMA) model and polynomial prediction model are established respectively. Secondly, the prediction values of each model act as the input of LS-SVM and the actual values serve as the output to form a combination prediction model. Trained by LS-SVM algorithm, the nonlinear combination model has good fitting effect and strong generalization ability. The proposed method is put into practical logistics demand prediction. The simulation results show the precision of the proposed model is higher than any of the single models and the average weights combination prediction model.