基于特征工程的常规公交车辆聚簇预测

The prediction of conventional buses gathering based on feature engineering method

有序、高效的运营是现代化公共交通系统应具备的主要特征之一。然而大量实测数据显示,公交线路上普遍存在着车辆偏离调度聚集行驶以及在站点处密集到达的聚簇现象,影响了乘客的出行体验,也造成了运力的浪费。本研究从数据驱动的角度出发,提出了基于特征工程的常规公交车辆聚簇行为预测的思路,并结合青岛市322路公交车实际运行数据完成了理论模型验证。结果表明,基于“本站前序班次的车头时距”以及“上游三个站点的车头时距的方差”两个特征的Logistic回归模型可以较好地预测公交车辆聚簇行为的发生,对于缓解公交车辆运行过程中的聚簇现象具有较强的现实意义。

Orderly and efficient operation is one of the main characteristics of a modern public transportation system.However,a large number of measured data show that there are generally clusters of vehicles on bus lines that deviate from the schedule and gather together and arrive densely at the station,which affects the travel experience of passengers and also causes a waste of transportation capacity.From a data-driven perspective,this study proposes an idea for predicting the clustering behavior of conventional public transport vehicles based on feature engineering,and completes the theoretical model verification based on the actual operation data of No.322 buses in Qingdao.The results show that the logistic regression model based on the two characteristics of"headway of the previous train at this station"and"variance of the headway of the three upstream stations"can better predict the occurrence of clustering behavior of public transport vehicles,and it is necessary to alleviate the The clustering phenomenon during the operation of public transport vehicles has strong practical significance.