Application of LSSVM to logistics demand forecasting based on grey relational analysis and kernel principal component analysis

Precise forecasting of logistics demand is very important for logistics system planning and designing. However, logistics demand is affected by many factors, which will cause the complex model for logistics demand forecasting. To simplify the forecasting model and improve the forecasting precision, this paper proposed a least squares support vector machines (LSSVM) model based on grey relational analysis (GRA) and kernel principal component analysis (KPCA) for forecasting logistics demand. Firstly, the GRA was applied determining the main factors affecting logistics demand. Secondly, the KPCA was applied to eliminating the correlation among the main factors and extracting the nonlinear principal components. Finally, the extracted nonlinear principal components were as the input variables to build LSSVM model for logistics demand forecasting. The parameters in LSSVM were optimized by the adaptive inertia weight particle swarm optimization (AIWPSO) algorithm. The logistics demand in China was used to evaluate the effectiveness of the proposed model. The results indicate that the proposed model greatly reduces the dimensions of the input variables and improves the forecasting precision for logistics demand.