Short-term inbound rail transit passenger flow prediction based on BILSTM model and influence factor analysis

Accurate and real-time passenger flow prediction of rail transit is an important part of intelligent transportation systems(ITS).According to previous studies,it is found that the prediction effect of a single model is not good for datasets with large changes in passenger flow characteristics and the deep learning model with added influencing factors has better prediction accuracy.In order to provide persuasive passenger flow forecast data for ITS,a deep learning model considering the influencing factors is proposed in this paper.In view of the lack of objective analysis on the selection of influencing factors by predecessors,this paper uses analytic hierarchy processes(AHP)and one-way ANOVA analysis to scientifically select the factor of time characteristics,which classifies and gives weight to the hourly passenger flow through Duncan test.Then,combining the time weight,BILSTM based model considering the hourly travel characteristics factors is proposed.The model performance is verified through the inbound passenger flow of Ningbo rail transit.The proposed model is compared with many current mainstream deep learning algorithms,the effectiveness of the BILSTM model considering influencing factors is validated.Through comparison and analysis with various evaluation indicators and other deep learning models,the results show that the R2 score of the BILSTM model considering influencing factors reaches 0.968,and the MAE value of the BILSTM model without adding influencing factors decreases by 45.61%.

Research on Railway Passenger Flow Prediction Method Based on GA Improved BP Neural Network

This paper chooses passenger flow data of some stations in China from January 2015 to March 2016, and the time series prediction model of BP neural network for railway passenger flow is established. But because of its slow convergence speed and easily falling into local optimal solution of the problem, we propose to improve the time series model of BP neural network by genetic algorithm to predict railway passenger flow. Experimental results show that the improved method has higher prediction accuracy and better nonlinear fitting ability.

Probabilistic interval prediction of metro-to-bus transfer passenger flow in the trip chain

To accurately analyze the fluctuation range of time-varying differences in metro-to-bus transfer passenger flows,the application of a probabilistic interval prediction model is proposed to predict transfer passenger flows.First,bus and metro data are processed and matched by association to construct the basis for public transport trip chain extraction.Second,a reasonable matching threshold method to discriminate the transfer relationship is used to extract the public transport trip chain,and the basic characteristics of the trip based on the trip chain are analyzed to obtain the metro-to-bus transfer passenger flow.Third,to address the problem of low accuracy of point prediction,the DeepAR model is proposed to conduct interval prediction,where the input is the interchange passenger flow,the output is the predicted median and interval of passenger flow,and the prediction scenarios are weekday,non-workday,and weekday morning and evening peaks.Fourth,to reduce the prediction error,a combined particle swarm optimization(PSO)-DeepAR model is constructed using the PSO to optimize the DeepAR model.Finally,data from the Beijing Xizhimen subway station are used for validation,and results show that the PSO-DeepAR model has high prediction accuracy,with a 90%confidence interval coverage of up to 93.6%.

The Research of Urban Rail Transit Sectional Passenger Flow Prediction Method

This paper studies the short-term prediction methods of sectional passenger flow, and selects BP neural network combined with the characteristics of sectional passenger flow itself. With a case study, we design three different schemes. We use Matlab to realize the prediction of the sectional passenger flow of the Beijing subway Line 2 and make comparative analysis. The empirical research shows that combining data characteristics of sectional passenger flow with the BP neural network have good prediction accuracy.

Prediction of Passenger Flow at Sanya Airport Based on Combined Methods

It is crucial to correctly predict the passenger flow of an air route for the construction and development of an airport.Based on the passenger flow data of Sanya Airport from 2008 to 2016,this paper respectively adopted Holt-Winter Seasonal Model,ARMA and linear regression model to predict the passenger flow of Sanya Airport from 2017 to 2018.In order to reduce the prediction error and improve the prediction accuracy at meanwhile,the combinatorial weighted method is adopted to predict the data in a combined manner.Upon verification,this method has been proved to be an effective approach to predict the airport passenger flow.

A prediction model to forecast passenger flow based on flight arrangement in airport terminals

Passenger flow plays an important role in the indoor environment and energy consumption of airport terminals.In this paper,field investigations were carried out in four typical airport terminals with different scales and operation states to reveal the characteristics of passenger flow.A prediction model is established to forecast passengers’distribution in the main areas of an airport terminal based on its flight arrangement.The results indicate the dislocation peaks of passenger numbers in these areas,due to the airport’s departure process.The peak time interval is about 30 min between the check-in hall and the security check area,and 60-80 min between the check-in hall and the departure hall.RD value(i.e.,the ratio of the actual passenger number in a certain area to the design value)is used to describe this peak shifting feature.When the annual passenger throughput of an airport terminal reaches or even exceeds its design value,the total peak RD value is normally 0.6-0.8.For the airport affected by COVID-19,the peak RD is only 0.2,which reflects the decline in terminal passenger numbers during the pandemic.This research provides useful insight into the characteristics of passenger flow in airport terminals,and is beneficial for their design and operation.

高速铁路日常客运量的EMD-Informer组合预测方法

铁路客流需求的科学预测是进行运输组织方案决策的重要依据。以高速铁路历史客票数据为基础,结合经验模态分解(empirical mode decomposition,EMD)与机器深度学习中的注意力机制,提出高速铁路日客流量的EMD-Informer组合预测方法。首先采用EMD方法分解高速铁路客流量序列,获得具有周期特征和线路客流内在特征的模态分量,再利用Informer模型分别训练和预测各模态分解分量,并通过多头注意力机制高效挖掘客流数据内在规律和捕捉数据序列中的关键特征,在此基础上重组各分量预测值,从而得到高速铁路日常客流的整体高精度预测值。同时,根据结合问题特征的大量实验,明确可供实际运用参考的超参数设置规则。基于京沪高速铁路全线的实例计算分析表明,相对对比预测方法,EMD-Informer组合预测方法在高速铁路客流的单步预测及超前预测上均具有明显更小的预测误差。

基于组合深度学习的轨道交通短时进站客流预测模型

针对轨道交通短时进站客流考虑不充分和特征学习不全面而导致预测精度不高的问题,选取客流特征、天气、空气质量和道路交通拥堵指数等多个因素,提出了一种基于组合深度学习的轨道交通短时进站客流预测模型(CNN-ResNet-BiLSTM)。基于卷积神经网络(CNN)对多因素客流时间序列进行自动提取,在CNN网络中插入多个残差神经网络(ResNet)来加深网络深度,利用双向长短时记忆神经网络(BiLSTM)捕捉前后两个方向的客流时间序列特征并得到预测结果;以杭州市全网80个站点工作日的进站客流为例,验证了该模型的有效性。研究结果表明:与常用的几种模型相比,多因素CNN-ResNet-BiLSTM组合模型的均方根误差(E RMS)至少降低了8.50%,平均绝对误差(E MA)至少降低了6.74%,平均绝对百分比误差(E MPA)至少降低了6.52%。

多尺度融合与动态自适应图的公交客流预测模型

公交客流预测是公共交通规划和管理中的重要问题。虽然时空图卷积在地铁客流预测任务中获得了很好的预测效果,但是面对公交更复杂的线路、大规模的节点数据,现有的基于图卷积的空间建模方法将带来巨大的空间内存消耗。同时,公交客流量短时间范围内更可能受到瞬时交通状况的影响。为了解决这些挑战,提出了一种多尺度融合和动态自适应图的公交客流预测模型(MFDAG)。该模型融合客流、时刻和周信息以增加数据的特征维度,用动态自适应图的方法来学习不同站点之间的关系。进一步提出了一种多尺度融合传播的方法来表示复杂的空间依赖关系,同时设计了一种多尺度卷积传播的方法来学习不同尺度的时间依赖关系。在两个真实的客流数据集上进行了实验,并与其他交通预测方法进行了比较。实验结果表明,所提出的多尺度融合和动态自适应图的公交客流预测方法具有更高的预测准确度。

基于迁移学习的轨道交通特殊OD客流预测研究

客流预测一直是轨道交通运营公司关注的重点,由于受到运输能力的限制等因素影响,部分OD的实际客流数据与真实需求有偏差,出现异常或者样本缺失,从而造成总体样本量偏小,直接采用这些样本进行预测会明显影响预测精度,但通过还原样本值增加样本量难度太大。根据上述特点选择基于实例的迁移学习,先确定源域的对象和范围,从源域中选择合适的样本补充到总体样本中,共同组成最终的训练样本数据集,完成迁移学习。同时选择改进的Boost算法,通过误差调整样本权重,不断迭代,得到最终的预测模型。结果表明:基于实例的迁移学习结合改进Boost算法的预测精度要好于传统集成学习、ARIMA模型、多元回归模型,为轨道交通运营公司对特定OD的客流预测提供新的有益尝试。

基于卷积长短时记忆网络的短时公交客流量预测

针对传统的短时客流预测方法没有考虑到时序特征中跨时段客流之间的相似性问题,提出一种改进k-means聚类算法与卷积神经网络和长短时记忆网络相结合的短时客流量预测模型k-CNN-LSTM。通过k-means算法对跨时段时序数据进行聚类,使用间隔统计确定k值,构建交通流矩阵模型,采用CNN-LSTM网络处理具有时空特征的短时客流。该模型能够对具有空间相关性的数据进行较为准确的预测。使用真实数据集对模型进行检验和参数调优,实验结果表明:k-CNN-LSTM模型较其他模型有相对较高的预测精度。

基于多维可预知特征的TCN-LSTM城轨短期客流预测

地铁客流量波动受众多因素影响,准确的客流预测数据有利于制定更高效的行车控制方案和客流管控方案。为提高客流预测精度,提出一种基于多维可预知特征的时序卷积神经网络-长短期记忆神经网络模型(TCNLSTM)地铁短期客流预测方法。考虑外部因素的影响,引入Spearman相关系数分析并提取日期、天气等可预知特征及其状态集,以提升预测精度,缩小特征空间,克服了冗余特征数据导致的模型过于复杂问题;通过融合时序卷积神经网络(TCN)提取的客流时间序列特征和可预知特征状态集构建了长短期记忆神经网络(LSTM)层输入,组合模型学习客流与外部影响因素的长短期依赖,从而实现常规日、节假日、不同天气等多场景下的短期客流预测。基于某西南城市地铁刷卡交易数据,对比差分整合移动平均自回归模型(ARIMA)、TCN、LSTM及TCN-LSTM模型的短期客流预测结果,得出组合模型的总体平均绝对误差(MAE)值比其他方法低27%~48%,均方误差(MSE)值低13%~35%,平均绝对百分比误差(MAPE)值低2.8%~6.7%,上述3项指标均表明TCN-LSTM模型的客流预测效果更好。此外,对比实验表明通过融入提取的可预知特征数据,TCN-LSTM模型在测试集上的预测误差评价指标明显降低,所提方法能有效提高地铁短期客流预测精度。

Probability tree based passenger flow prediction and its application to the Beijing subway system

In order to provide citizens with safe, convenient and comfortable services and infrastructure in a metropolis, the prediction of passenger flows in the metro-net of subway system has become more important than ever before. Al- though a great number of prediction methods have been pre- sented in the field of transportation, all of them belong to the station oriented approach, which is not well suited to the Bei- jing subway system. This paper proposes a novel metro-net oriented method, called the probability tree based passenger flow model, which is also based on historic origin-destination （OD） information. First it learns and obtains the appearance probabilities for each kind of OD pair. For the real-time origin datum, the destination datum is calculated, and then several kinds of passenger flow in the metro-net can be pre- dicted by gathering all the contributions. The results of exper- iments, using the historical data of Beijing subway, show that although the proposed method has lower performance than existing prediction approaches for forecasting exit passenger flows, it is able to predict several additional kinds of passen- ger flow in stations and throughout the subway system; and it is a more feasible, suitable, and advanced passenger flow prediction model for Beijing subway system.

基于可预知大客流的城市轨道交通列车加开计划调整模型与算法

[目的]节假日、大型文体活动等事件会导致城市轨道交通线路产生可预知大客流,需在确保列车运行安全的前提下,根据客流预测情况在已有列车运行图中加开列车,以期在运营成本、运营组织复杂度和乘客服务水平之间取得平衡。[方法]采用事件-活动网络,对列车运行过程进行了形式化描述,建立了3个目标函数及相关的约束条件,进而构建了独立运用策略下列车加开计划调整模型、套跑混合运用策略下列车加开计划调整模型。在此基础上,提出了基于混合遗传禁忌的搜索算法。结合深圳地铁11号线的实际运营数据,在高峰时段和非高峰时段分别建立了不同的算例场景,以验证所建模型和算法的有效性。[结果及结论]所建模型支持不同运营时段、不同运行交路、不同车底接续策略下列车加开需求,在可预知大客流的运营调整上具有适用性。在车辆资源相对有限的情况下,为有效缓解高峰时段的客流压力,可灵活采用大小交路套跑混合运用策略,使线路的运输能力最大化。

基于异构数据特征的城市轨道交通OD客流短时预测方法

城市轨道交通起讫点(origin-destination,OD)客流短时预测在智能交通系统中意义重大,它为交通管控策略实施以及出行者出行选择提供了重要的决策依据。卷积神经网络被广泛用于交通数据空间相关性提取,但其平移不变性与局部敏感性导致该方法更重视局部特征而忽视全局特征。本研究构建了基于注意力机制的异构数据特征提取机模型(heterogeneous data feature extraction machine,HDFEM)以实现OD矩阵空间相关性的全局感知。该模型从时空特征和用地属性特征出发,构造异构数据OD时空张量与地理信息张量,依托模型张量编码层对异构数据张量进行分割与编码,通过注意力机制连接各张量块特征,提取OD矩阵中各个部分间的空间相关性。该方法不仅实现了异构数据与OD客流数据的融合,还兼顾了卷积神经网络模型未能处理的OD矩阵远距离特征,进而帮助模型更全面地学习OD客流的空间特征。对于OD时序特性,该模型使用了长短时记忆网络来处理。在杭州地铁自动售检票系统(auto fare collection,AFC)数据集上的实验结果表明:HDFEM模型相对于基于卷积神经网络的预测模型,其均方误差、平均绝对误差与标准均方根误差分别下降了4.1%,2.5%,2%,验证了全局OD特征感知对于城市轨道交通OD客流预测的重要性。

基于时序聚类CEEMDAN-LSTM的轨道短时客流预测

目的:高效精确的短时客流预测是城市轨道交通运营管理的重要前提,为提高短时客流预测精度,提出一种基于时序聚类的CEEMDAN-LSTM组合模型。方法:以DTW距离为度量标准,采用Kmeans算法对客流时序进行分类,在此基础上通过CEEMDAN算法进行时序分解以弱化样本噪声干扰,再将分量输入到LSTM模型中进行预测。结果:CEEMDAN-LSTM模型在3类客流时序下的预测误差均小于其他4个基线模型,并能有效反映短时客流的变化趋势;考虑时序聚类的预测模型的预测精度与时效性均优于不分类下的预测模型。结论:以合肥南站地铁的短时进站客流数据为例进行实证分析,证实客流时序聚类对预测精度提升的贡献,并与SARIMA、RF、XGBoost、LSTM等4个预测模型比较,CEEMDAN-LSTM模型具有较高的预测精度,且能有效反映实际客流曲线的变化趋势。

基于改进LSTM的城市轨道短时流量预测研究

针对LSTM应用于城市轨道短时流量预测存在的模型参数确定困难、对预测精度影响大的问题,采用改进的BA算法对模型参数进行优化。对传统BA算法,采用自适应策略来动态调整脉冲频率和蝙蝠速率,同时蝙蝠位置更新模型中引入随机扰动项,提高了BA的优化性能。采用BA对LSTM参数进行优化,提出了基于改进LSTM的城市轨道短时流量预测模型。将提出的模型应用于郑州地铁1号线,通过和BP神经网络预测模型、LSTM预测模型的对比,验证了所提出的改进LSTM预测模型具有更高的预测精度。

基于时空关联的时空图卷积神经网络城市轨道交通进站客流预测

[目的]准确的短时客流预测对于提高超大规模城市轨道交通线网的运营和管理效率具有重要意义,而目前现有研究对于深度挖掘时空关联性仍不够充分,为此基于短时客流的时空规律提出了基于客流时空关联特征的STGCN(时空图卷积神经网络)模型。[方法]首先,通过切比雪夫图卷积网络捕捉超大规模城市轨道交通网络的空间相关性,借助门控循环单元挖掘多时空关联特征下客流的时间相关性;然后,分析待预测车站历史客流数据相关性及OD(起讫点)客流数据相关性,以深入提取时空相关性;最后,结合客流时空关联特征建立STGCN模型。[结果及结论]以上海地铁江苏路站为例,进行短时进站客流预测,结果表明采用时空关联特征参数的预测结果较未加入特征参数的预测精度提高了16%,预测效果较优。

城市轨道交通乘客画像构建方法及应用策略研究

用户画像广泛应用于个性化推荐、广告精准投放等领域,但目前轨道交通领域的相关研究尚不完善,存在乘客画像的信息少、维度单一、指标设计和挖掘不充分等问题。通过从运营管理者和乘客服务多角度系统梳理互联网+城市轨道交通乘客画像的新需求,进而结合AFC和土地等多源数据,构建满足新需求下的轨道交通乘客画像指标体系,并给出关键性指标的计算和推断方法。以北京地铁为例,构建并验证轨道交通乘客画像构建方法的有效性,分析了乘客画像在辅助精细化客流预测方面的应用效果。结果表明,将反映乘客个体出行特征的返程客流加入到S-ARIMA模型中,相比S-ARIMA模型均方根误差(RMSE)下降9.02,对称平均绝对百分比误差(SMAPE)下降0.16%,有效提高客流预测精度。乘客画像未来还能应用在更多的场景,支撑轨道交通的智慧化、一体化出行服务。