摘要
为统筹安排运输资源、合理制定多式联运计划,针对集装箱多式联运货运量预测准确性不足的问题,基于网格搜索交叉验证改进的随机森林算法,构建一种集装箱多式联运月度货运量预测方法。首先对于原始数据特征中存在数据缺失的问题,采用二次插值法、K邻近插补法进行数据的填补;再利用平均准确度下降法和平均边际贡献法筛选出7个最优特征;最后构建基于网格搜索交叉验证的随机森林预测模型。以营口至武汉的多式联运案例进行验证,证明综合特征筛选与最优参数组合提升了模型预测的准确性以及K邻近插补法在模型中的优越性,并与三次指数平滑法、Xgboost模型进行对比,结果表明,改进的随机森林预测模型准确性和稳定性更高。
In order to make arrangements for transportation resources and formulate multimodal transport plans rationally, aiming at insufficient accuracy of container multimodal transport freight volume prediction, a method for container multimodal transport monthly freight volume prediction was constructed by using improved random forest algorithm based on grid search cross validation.First, for the missing data in the original data features, the quadratic interpolation method and the K-neighbor interpolation method were used to fill in the features.Then the mean decrease in accuracy method and the shapley additive explanation method were used to screen out 7 optimal features.Finally,the random forest prediction model based on grid search cross validation was established.The multimodal transport case from Yingkou to Wuhan was taken for verification, it proved that the comprehensive selection of features and the combination of optimal parameters have improved the accuracy of the model prediction and the superiority of K-neighbor interpolation in the model, then compared with the cubic exponential smoothing method and the Xgboost model, the results show that the improved random forest has higher accuracy and stability.
作者
辜勇
杨泽昭
GU Yong;YANG Ze-zhao(School of Transportation and Logistics Engineering,Wuhan University of Technology,Wuhan 430063,China)
出处
《武汉理工大学学报》
CAS
2023年第1期35-44,共10页
Journal of Wuhan University of Technology
基金
国家重点研发计划(2021YFB2601605)。
关键词
多式联运
集装箱货运量
网格搜索
交叉验证
随机森林
multimodal transport
container freight volume
grid search
cross validation
random forest
