摘要
为提升城市轨道交通站点短时客流预测精度,探讨其高峰客流集中,客流、区域人口密度间存在时空关联的特点,并针对性地建立了一种基于深度时空网络的城市轨道交通网路尺度下的站点客流短时预测模型。模型用卷积神经网络捕捉轨道站点之间的空间关联,融合了热点区域人数等因素作为外部特征,再用长短期记忆网络提取时间序列中的时间关联。选取深圳市轨道交通网络作为案例,基于深圳地铁客流数据,对全部站点的进、出站客流进行预测。通过与四种常用预测方法比较,验证了该方法的有效性和精度。
To improve the accuracy of short-term forecast of metro station passenger flow,this paper discussed its highly concentrated and spatial-temporal dependent characteristics,showing passenger flow is relevant to region population.a forecasting model based on deep spatial-temporal network is developed accordingly.Convolutional neural network(CNN)is applied to capture the spatial dependency between stations.Long short-term memory network(LSTM)is applied to capture the time dependency in the time series.Further,population in region of interests and holidays are incorporated as external features.Experiments on Shenzhen Metro well demonstrate the effectiveness and precision of the proposed model against 4 common baselines.
作者
黎旭成
彭逸洲
吴宗翔
陈振武
LI Xucheng;PENG Yizhou;WU Zongxiang;CHEN Zhenwu(Shenzhen Urban Transport Planning Center Co.,Ltd,Shenzhen 518000,China)
出处
《交通与运输》
2020年第S02期55-61,共7页
Traffic & Transportation
关键词
短时客流预测
城市轨道交通
时空关联分析
Short-term passenger flow forecast
Urban rail transit
Spatial-temporal dependency analysis
