基于RBF神经网络的区域物流需求预测

Regional Logistics Demand Forecasting Based on RBF Neural Network

区域物流需求预测是政府部门进行物流规划、制定物流相关政策的基础,具有数据量小和非线性的特点,传统预测方法难以保证预测的精度。本文将区域外贸总额、人均消费水平、区域零售总额和三大产业的产值作为影响因素,运用RBF神经网络和随机森林算法对区域物流需求进行预测,并结合多元线性回归、MLP神经网络、灰色预测模型、BP神经网络和基于泊松分布的神经网络算法进行对比分析。实例分析的结果表明,基于RBF神经网络的区域物流需求预测模型精度最高,平均绝对误差百分率为2.41%,最大绝对误差百分率为8.56%,达到了精准预测的效果,具有一定的应用价值。

Regional logistics demand forecasting is the basis for government departments to make logistics planning and related policies.The basic data shows the characteristics of small quantity and nonlinearity,so it is difficult to ensure the accuracy of forecasting by traditional forecasting methods.Based on influence factors including regional foreign trade total,per capita consumption level,regional retail sales and output value of the three major industries,we proposed to use RBF neural network and random forest to predict regional logistics demand.Multiple linear regression,MLP neural network,grey prediction model,BP neural network and neural network based on Poisson distribution were used for comparative analysis.The results show that with an average percentage of absolute error of 2.41%and the max percentage of absolute error of 8.56%,RBF neural network has a higher prediction accuracy,which displays certain practical application value.