基于AOA-LSSVM模型的枢纽城市物流需求量预测

Prediction of Logistics Demand in Hub Cities Based on AOA-LSSVM Model

传统的LSSVM难以全面反映物流需求的变化规律,会导致预测效果不佳。首先利用灰色关联分析(GRA)得到物流需求的主要影响因素;将主要影响因素作为LSSVM的输入变量,构建物流需求预测模型;通过阿基米德算法(AOA)对最小二乘支持向量机的正则化参数(γ)和核参数(σ)进行迭代寻优,以减少参数选择的盲目性;构建AOA算法优化最小二乘支持向量机(LSSVM)的智能预测模型AOA-LSSVM,经过验证该模型可以提高预测精度。运用AOA-LSSVM模型对西部陆海新通道的重要枢纽城市——重庆、成都、贵阳和南宁的物流需求进行实证分析,结果表明:该模型与LSSVM模型相比取得较高的预测精度,其均方根误差、平均绝对误差、以及异方差调整的均方根误差、异方差调整的平均绝对误差分别降低了1946.4,1206.1,0.0284,0.0397。

Traditional LSSVM(least squares support vector machine)is difficult to comprehensively reflect the change pattern of logistics demand,which will cause poor prediction performance.Grey correlation analysis(GRA)was used to obtain the main influencing factors of logistics demand,and then the main influencing factors were used as input variables of LSSVM to construct a logistics demand forecasting model.The regularization parameters(γ)and kernel parameters(σ)of LSSVM were iteratively optimized by the Archimedes algorithm(AOA)to reduce the blindness of parameter selection.Thus,an intelligent prediction model AOA-LSSVM was constructed to optimize LSSVM by using the AOA algorithm.After verification,the proposed model can improve prediction accuracy.The proposed AOA-LSSVM model was used to empirically analyze the logistics demand of the important hub cities of the western land and sea new passage such as Chongqing,Chengdu,Guiyang and Nanning.The results show that compared to the LSSVM model,the proposed model can achieve higher prediction accuracy,with a reduction of 1946.4,1206.1,0.0284 and 0.0397 in root mean square error,mean absolute error,and heteroskedasticity-adjusted root mean square error and mean absolute error,respectively.