利用标准差赋权组合模型预测大宗货物运输需求  

Forecasting the transportation demand of bulk goods by using the weighted combination model of standard deviation

作　　者：崔淑华[1] 侯慧君 武慧荣[1] CUI Shuhua;HOU Huijun;WU Huirong(College of Transportation,Northeast Forestry University,Harbin 150040,China)

机构地区：[1]东北林业大学交通学院,哈尔滨150040

出　　处：《交通科技与经济》2022年第6期31-38,共8页Technology & Economy in Areas of Communications

基　　金：2020年黑龙江省交通运输厅科技项目;中央高校基本科研业务费专项资金项目(2572015CB16)。

摘　　要：目前关于运输需求预测研究,大多采用以历史运输量数据为基础进行直接预测的方法,但这种使用单一模型且不考虑关联因素的预测方法,易导致预测结果与实际值偏差较大。提出基于标准差赋权,建立多元线性回归预测模型与GM(1,1)-MLP神经网络预测模型并联结构的组合预测模型。以哈尔滨市粮食产量的历史数据验证组合预测模型的有效性,结合产运系数,对目标年哈尔滨市粮食运输需求进行预测,为大宗货物运输组织方案设计提供数据支持。结果表明:相较于两种单一预测模型,组合预测模型的预测精度更高,能反映哈尔滨市的粮食产量、运输需求量及变化趋势。At present, most of the research on forecasting of transportation demand adopts the method of direct forecasting based on historical transportation volume data. However, this forecasting method that uses a single model and does not consider related factors easily leads to a large deviation between the forecasting results and the actual value. Based on the standard deviation weighting, a combined forecasting model with a parallel structure of multiple linear regression forecasting model and GM(1,1)-MLP neural network forecasting model is proposed. The validity of the combined prediction model is verified by the historical data of grain output in Harbin, combined with the production and transportation coefficient, to forecast the grain transportation demand in Harbin in the target year, and provide data support for the design of bulk cargo transportation organization plan. The results show that compared with the two single forecasting models, the combined forecasting model has higher forecasting accuracy and can reflect the grain output, transportation demand and changing trend of Harbin.

关 键 词：货物运输 运输需求 组合预测模型 多元线性回归模型 GM(1 1)-MLP神经网络模型 产运系数 

分 类 号：U492.313[交通运输工程—交通运输规划与管理]