基于PCA-PSO-RBF模型的水产品冷链物流需求预测-《福建工程学院学报》

文章信息/Info

Title:: Prediction of cold chain logistics demand for aquatic products based onPCAPSORBF model

Author(s):: ZHUANG Weiqing1; WANG Haoran; School of Internet Economics and Business, Fujian University of Technology

Keywords:: aquatic products; demand prediction; principal component analysis; particle swarm algorithm; radial basis function neural network

摘要:: 为确保水产品冷链物流供需双方信息对称,降低供应链中断风险及供需不匹配造成的浪费,水产品冷链物流需求预测显得尤为关键。选取影响水产品冷链物流需求的18个因素并用灰色关联法(GRA)筛选验证,运用主成分分析法(PCA)提取主要特征,通过粒子群算法(PSO)优化的径向基神经网络(RBF)构建PCA?PSO?RBF预测模型,对水产品需求预测,并与PCA?PSO?BP、PCA?RBF、PCA?BP、SVM、BP模型对比。结果表明,构建的PCA?PSO?RBF预测模型具有较强的非线性系统处理能力与全局寻优能力,对小样本多特征的数据具有较好包容性和预测精度,通过MSE/RMSE/MAPE预测误差评价验证了PCA?PSO?RBF预测模型的有效性及优越性。

Abstract:: In order to ensure the information symmetry between supply and demand sides of cold chain logistics of aquatic products, reduce the risk of supply chain disruption and the waste caused by the mismatch between supply and demand, the prediction of demand for aquatic cold chain logistics is particularly critical. Firstly, 18 factors affecting the demand of cold chain logistics for aquatic products were selected and validated by screening with gray correlation method (GRA). Then the main features were extracted by using principal component analysis (PCA), and a radial basis function neural network (RBF) optimized by particle swarm algorithm (PSO) was used to construct a PCAPSORBF prediction model to forecast the demand of aquatic cold chain logistics, and the model was compared with PCAPSOBP, PCARBF, PCABP, SVM, and BP models for comparison study. Results show that the constructed PCAPSORBF prediction model has strong nonlinear system processing ability and global optimization ability, and has good inclusiveness and prediction accuracy for data with small samples and multiple features. The effectiveness and superiority of PCAPSORBF prediction model were verified by MAE/RMSE/MAPE prediction error evaluation.