基于时空特征提取的城市轨道交通乘客出行目的地预测

为满足城市轨道交通运营组织进行客流管控和行车调度的实时需求，提出了基于乘客OD时空特征的出行目的地在线预测方法。通过分析定义乘客OD时空特征矩阵，以乘客个体的历史自动售检票系统（AFC）数据为训练样本，提出了基于行程密度聚类的乘客OD时空特征提取方法。分析制定乘客实时进站刷卡信息与其OD时空特征矩阵的匹配规则，基于3种匹配情况分别提出了相应的目的地实时预测方法。以南京市轨道交通AFC数据为实例进行验证，结果表明本文提出的预测方法在高峰时段预测准确率、全天预测稳定性等方面效果良好，可为地铁运营组织提供参考。

Travel destination prediction method for urban rail transit passengers using spatiotemporal feature extraction

To meet the real-time demand for passenger flowcontrol and train dispatching in urban rail transit operation organizations, anonline travel destination prediction method based on passenger origindestination(OD) spatiotemporal characteristics is proposed. First, thepassenger OD spatiotemporal feature matrix is defined. Moreover, the automaticfare collection (AFC) history data of individual passengers are used astraining samples, and an extraction method of passenger OD spatiotemporalfeatures is proposed using stroke density clustering. Further, based on thematching rule analysis between the card-swiping information of passengers andtheir OD spatiotemporal feature matrix, the corresponding real-time destinationprediction method is proposed for three matching cases. The AFC data of Nanjingrail transit are used to verify the proposed method. Results show that theproposed method is effective in the prediction accuracy of rush hours andprovides stable all-day prediction, which can provide a reference for subwayoperations.