物流需求预测方法研究进展

Research Progress of Methods of Logistics Demand Forecasting

物流需求的准确预测对于物流发展政策的制定、物流基础设施的确定、物流市场态势的分析、物流资源利用率的提高等方面具有重要的理论意义和实际应用价值。从建模形式出发,将已有物流需求预测方法分为单一传统预测方法、单一智能预测方法、组合预测方法、混合预测方法四大类。其中,单一传统预测方法主要包括单纯的时间序列法、回归分析、数理统计方法等,单一智能预测方法主要涉及灰色预测法、神经网络、支持向量机以及它们的改进形式;组合预测方法主要归纳为三种组合形式:线性组合单一预测结果、非线性组合单一预测结果、修正单一预测结果;混合预测方法主要总结为三种混合形式:混合智能优化算法与单一预测方法、混合数据降维技术与智能预测方法、混合数据挖掘技术与智能预测方法。综述了四大类预测方法,从建模原理、优缺点及适用性等方面对四大类方法中的各预测模型进行评析,以期为物流需求研究人员寻到适合于不同物流需求预测任务的预测方法。

Starting from the form of modeling,we divide the existing logistics demand forecasting methods into four categories,namely single traditional forecasting methods,single intelligent forecasting methods,combined forecasting methods,and hybrid forecasting methods.Among them,the single traditional prediction methods mainly include such simple methods as the time series method,regression analysis method,and mathematical statistics method,etc.;the single intelligent prediction methods mainly involve the gray prediction method,neural network method,support vector machine method,and their improved forms;the combined prediction methods mainly fall into three groups,namely the linear combination single forecasting method,the nonlinear combination single forecasting method,and the modified single forecasting method;and the hybrid forecasting methods mainly fall into three mixing forms,namely the hybrid intelligent optimization algorithm and single forecasting method,the hybrid data reduction technology and intelligence forecasting method,and the hybrid data mining technology and intelligent forecasting method.Next,we review the four major types of forecasting methods,and assess the forecasting models in the four major categories in terms of modeling principles,advantages and disadvantages,and applicability,with a view to finding the suitable method for the different logistics demand forecasting tasks.