A Regional Logistics Demand Forecasting Method using KPCA-GA-ELM

Abstract

Logistics demand forecasting works as a basis for well operated city governance especially for e-commerce industry. Yet, how to accurately predict the local logistics demand remains further improvement. To deal with it, this paper proposed an KPCA-GA-ELM approach which firstly introduces an extreme learning machine (ELM) approach to build the forecasting model, then incorporate both kernel principal component analysis (KPCA) and genetic algorithm (GA) into it. Taking Shanghai's regional logistics demand prediction as an example, two principal components affecting regional logistics demand are extracted by KPCA, ELM is then used to develop a regional logistics demand forecast model, and the genetic algorithm was applied to make the ELM model arguments be better to avoid the impact of strong randomness in parameter selection on model prediction performance and generalization ability. The results indicate that the accuracy is significantly improved comparing it with other two models. Such model then can be used as the demand forecasting and estimation approaches to estimate the demand of other industries in a region.