Abstract: Carpooling is an environmentally friendly and energy saving way to travel. Excellent carpooling strategy can not only relieve the traffic pressure, optimize passengers’ experience, but also reduce carbon emissions and so on. To solve carpooling problem, this paper employs a two-stage carpool matching strategy. In the first stage, this paper employs the cluster method based on Hausdorff distance to assign travel demands to specific vehicle. And in the second stage, the cluster method based on matching degree is proposed to choose the most appropriate travel demands for every car. The experimental result shows that this method has achieved the goal to make recommendations for both single-vehicle carpooling and multi-vehicle carpooling.

Key words: data mining, clustering, multi-carpooling, Hausdorff distance, matching degree

摘要： 拼车是一种环保节能的出行方式，合理的拼车策略可以缓解交通压力，优化乘客体验，减少碳排放等。针对拼车问题，提出了两阶段的拼车匹配策略。第一阶段匹配过程是利用基于改进Hausdorff距离的乘客分配算法，将拼车需求分配到具体车辆，从而将多车辆问题转化为单车辆问题；第二阶段匹配过程，采用基于匹配度的聚类筛选出与车辆最为匹配的拼车需求。实验结果表明该算法和流程能分别应用于单车次、多车次接力换乘的拼车方案推荐，匹配简单准确。

关键词: 数据挖掘, 聚类, 接力拼车, 豪斯多夫距离, 匹配度