为给车辆调度优化以及集卡预约系统设计提供重要参考,利用所采集的深圳市某港口的码头闸口数据，建立一种基于数据挖掘的集卡周转时间短时预测方法.首先，通过对码头闸口数据进行分析，获取车辆到达时间分布、任务类型、作业方式等集卡作业特征以及集卡在码头内的周转时间；在此基础上，利用循环神经网络并结合训练集数据，建立集卡作业特征与其周转时间之间的映射关系.其次，为减少随机波动对周转时间预测效果的影响，利用小波分解算法对循环神经网络拟合结果的残差进行高频噪声分离，并通过自回归模型拟合过滤后的低频序列.最后，将拟合后的循环神经网络与自回归模型进行结合，建立一种支持集卡周转时间短时预测的组合模型，并利用测试集数据进行有效性验证.结果表明，相比单一的循环神经网络，该组合模型可以大幅提升预测精度.

In order to provide an important reference for vehicle scheduling optimization and truck reservation system design, a shortterm prediction method of truck turnover time based on data mining was established by using the collected data of a port in Shenzhen. Firstly, by analyzing the data of the terminal gate, the truck operation characteristics such as vehicle arrival time distribution, task type, the operation mode and the turnover time of the truck in the terminal were obtained. On this basis, the mapping relationship between truck operation characteristics and turnover time was established by using recurrent neural network (RNN) and training set data. Secondly, in order to reduce the impact of random fluctuations on the prediction of turnaround time, the wavelet decomposition algorithm was used to separate the residual of the fitting results of the recurrent neural network with highfrequency noise, and the filtered lowfrequency series was fitted by the autoregressive model. Finally, the combined model of the fitted cycle neural network and the autoregression model (AR) was established to support the shortterm prediction of the truck turnaround time, and verified by test set data. The results show that the combined model can greatly improve the prediction accuracy compared with the single RNN.