摘要
针对网络货运平台的异常轨迹检测问题,提出了基于循环神经网络的异常轨迹跟踪检测策略。首先采用基于循环神经网络的轨迹嵌入方法,学习轨迹的序列信息。该轨迹嵌入能刻画异常轨迹和正常轨迹之间的内部特征。随后提出了基于注意力的机制,实现长序列信息的聚合,以进一步提高轨迹嵌入的质量。最后,使用真实数据集对提出的策略进行实验评估。结果表明,与现有方法相比,本文提出的策略具有较好的异常轨迹检测性能。
Aiming at the problem of abnormal trajectory detection on the network freight platform,this paper proposes an abnormal trajectory tracking and detecting strategy based on recurrent neural network.First,a trajectory embedding method based on recurrent neural network is applied to learn the sequence information of the trajectory.The trajectory embedding can describe the internal characteristics between abnormal trajectory and normal trajectory.Then,an attention-based mechanism was subsequently proposed to aggregate long sequence information,to improve the quality of trajectory embedding.Finally,the proposed strategy is evaluated experimentally by using real data sets.The results show that,compared with the existing methods,the strategy proposed in this paper has better anomalous trajectory detection performance.
作者
齐晗
QI Han(School of Economics and Management,Anhui Vocational College of Electronics&Information Technology,Bengbu 233000,China)
出处
《信阳农林学院学报》
2022年第3期102-106,共5页
Journal of Xinyang Agriculture and Forestry University
基金
2022年安徽省高校优秀青年人才支持项目(gxyq2022266)
安徽省人文社科重点项目(SK2021A1062)
关键词
异常轨迹检测
网络货运平台
循环神经网络
注意力机制
abnormal trajectory detection
network freight platform
recurrent neural network
attention mechanism
