基于时空聚类的定制公交需求响应机制

Demand Response Mechanism of Customized Bus Based on Space-time Clustering

预约需求的响应对定制公交的运营至关重要,但是乘客的预约需求是时空分散的,运输企业往往靠经验来确定是否响应预约需求,很有可能降低定制公交的吸引力。提出了一种基于时空聚类法的定制公交需求响应机制。通过时空维度的响应对预约需求点进行筛选,首先从时间维度进行筛选,采用基于时间度量的层次聚类算法保留与出行时间接近的预约需求点;然后从空间维度进行筛选,运用DBSCAN聚类算法剔除空间位置相对孤立且人数较少的特殊请求,得到时空趋同的大众化需求。为了验证该响应机制的有效性,进行了算例分析。结果表明,在聚类参数按照经验值设定的前提下,仅能够响应69%的预约需求点和75%的乘客;通过对聚类参数进行适当调整,当保留时间跨度的最短时间不大于3 min,保留预约需求点的最少乘客人数不大于2人,满足到达地位置接近的条件值不小于1000 m,DBSCAN聚类算法的输入参数邻域不小于400 m,聚为一类的乘客人数下限值不大于4人时,能够响应75%的预约需求点和80%的乘客,满足了尽量响应大部分定制需求、适当剔除特殊需求的响应原则。可见,该机制对定制需求的响应具有良好的适用性,可为运输企业开通定制公交线路提供决策依据。

The response of reserved demand is crucial to the operation of customized bus,however,the reserved demand of passengers is scattered in time and space.Transport enterprises often rely on experience to determine whether to respond to the reserved demand,and it is likely to reduce the attractiveness of customized bus.The customized bus demand response mechanism based on space-time clustering is proposed.The reserved demand stations are filtered by response on space-time dimension.First,the filtering is conducted in time dimension,and the reserved demand stations close to the travel time are retained by using hierarchical clustering algorithm based on time measurement.Then,the filtering is conducted in space dimension,the special requests with relatively isolated spatial location and fewer people are eliminated using the DBSCAN cluster algorithm,so as to obtain the popularization request with the convergence of time and space.In order to verify the effectiveness of the response mechanism,the case study is performed.The result shows that(1)on the premise of setting the clustering parameters according to the empirical value,only 69%of reserved demand and 75%of passengers can be responded;(2)by appropriately adjusting parameters,when the minimum retention time span is not more than 3 min,the minimum passengers to get reserved demand station is not greater than 2 persons,the condition value of approaching destination is not less than 1000 m,the neighborhood of input parameter of DBSCAN clustering algorithm is not less than 400 m,and the minimum passengers clustered into a group is not more than 4 persons,it can respond to 75%of the reserved demand stations and 80%of the passengers,meeting the response principle of responding to most of the customized demands as far as possible and properly eliminating special demands.It can be seen that the mechanism has great applicability to response customized demands,and can provide decision-making basis for transport enterprises to open customized bus lines.