基于集成学习的多式联运货运量预测模型构建

货运量精准预测是多式联运网络高效协同发展的重要基础,货运量时变性强、数据多样性缺失是实现精准货运量预测的问题所在。基于此,通过挖掘货物运输量(集装箱)的时间变化特征,构建初始相关时间特征输入集,结合斯皮尔曼相关性系数分布,采用Bagging+BP集成学习方法训练多个弱分类器,最终组合获取高精度的强学习模型。以南京龙潭港为例,对自回归移动平均模型(ARIMA)、Bagging+BP集成学习网络以及长短时记忆神经网络(LSTM)三种模型进行评价,实验结果表明,相比于其他模型,提出的Bagging+BP集成学习网络预测性能良好,有一定的实用价值。

Construction of Multimodal Transport Freight Volume Forecast Model Based on Ensemble Learning

In this paper,in light of the fact that accurate freight volume forecasting is an important foundation for the efficient and coordinated development of multimodal transport networks and that the strong time-varying nature of freight volume and the lack of data diversity are hindering the accurate forecasting of freight volume,and by uncovering the time varying characteristics of cargo transportation(container)volume,we constructed the time feature input set initially relevant,next,combining with the Spearman correlation coefficient distribution,used the Bagging+BP ensemble learning method to train multiple weak classifiers,and finally obtained a high-precision strong learning model.Then,taking Nanjing Longtan Port as an example,we reviewed the three models being autoregressive moving average model(ARIMA),Bagging+BP ensemble learning network and long-short-term memory neural network(LSTM),the result of which showed that compared with the other models,the Bagging+BP ensemble learning network proposed in this paper has good forecasting performance and is of certain practical value.