基于Attention⁃LSTM神经网络的公交行程时间预测  

Bus travel time prediction based on Attention⁃LSTM neural network

作　　者：徐丸絮 沈吟东[1] XU Wanxu;SHEN Yindong(School of Artificial Intelligence and Automation,Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]华中科技大学人工智能与自动化学院,湖北武汉430074

出　　处：《现代电子技术》2022年第3期83-87,共5页Modern Electronics Technique

基　　金：国家自然科学基金项目(71571076)。

摘　　要：传统的公交行程时间预测模型由于忽略了历史时刻中的信息,导致预测精度不理想。针对公交行程时间的时序性,提出一种基于LSTM神经网络的预测模型,并引入注意力(Attention)机制对其进行优化。首先,综合考虑多种影响因素,设计了多变量LSTM模块,将当前时刻的行程时间与历史时刻数据相关联,对其中的多维度特征进行信息提取;随后针对单一LSTM网络无法自动识别不同信息重要性的局限性,引入Attention机制,使模型聚焦重点信息、忽略冗杂信息;最后,采用实际公交GPS数据验证了该方法的有效性。实验结果表明,与五种常见方法相比,该模型具有更高的精度。The information in the historical moment is ignored in the traditional bus travel time prediction model,which results in unsatisfactory prediction accuracy.Therefore,a prediction model based on LSTM(long short⁃term memory)neural network is proposed according to the time sequence of travel time data.Attention mechanism is introduced to optimize the model.Various factors are taken into account comprehensively to design a multi⁃variable LSTM module,which associates the travel time at the current moment with the historical time data.The multi⁃dimensional feature information are extracted.Subsequently,in view of the limitation that a single LSTM network fails to automatically identify the importance of different information,the Attention mechanism is introduced to make the model focus on key information and ignore redundant information.Finally,the actual GPS data of the bus is used to verify the effectiveness of the model.The experimental results show that the proposed model has higher accuracy in comparison with the five common methods.

关 键 词：智能交通 公交行程时间预测 LSTM神经网络 Attention机制 公交GPS数据 深度学习 循环神经网络 

分 类 号：TN99-34[电子电信—信号与信息处理]