PTRA:一个面向空载出租车的路线推荐算法-《计算机技术与发展》

文章信息/Info

Author(s):: LI Si-jia; SU Fan-jun; School of Optical-electrical and Computer Engineering,University of Shanghai forScience and Technology,Shanghai 200093,China

Keywords:: taxi; maximizing revenue; real-time road conditions; recommendation; clustering

摘要:: 目前针对出租车的推荐系统主要从降低空载率和减少寻客里程两个方面为司机推荐潜在载客点或最优行驶路线,然而从司机收益最大化的角度而言,多数研究没有考虑实时路况的变化对推荐效果的影响。因此,以收益最大化为目标进行了研究,提出了一种空载出租车推荐算法 PTRA(profit-based taxi recommendation algorithm)。该算法包含两个阶段,在离线挖掘阶段,通过净利润目标函数用于评估驾驶路线的潜在利润,并通过 DBSCAN 聚类高收益路段区域;而在线推荐阶段,根据出租车当前位置结合当前路段反馈的实际路况为出租车司机提供高收益路线。该框架尤其对特殊情况(天气、演出等大型事件)敏感,可对热门区域路线进行多次推荐。实验采用海口市一个月的出租车 GPS 记录数据,最终实验结果验证了 PTRA 推荐算法的正确性和有效性。

Abstract:: Nowadays,the route recommendation system mainly recommends potential loading points or optimal driving routes for drivers from two aspects: reducing the idle rate and reducing the of passengers seeking miles. However,from the perspective of maximizing the driver’s revenue, most research do not concern the impact of changes in real - time road conditions on the effectiveness of recommendations. Therefore,in the context of maximizing revenue,a profit-based taxi recommendation algorithm is proposed called PTRA (profit-based taxi recommendation algorithm),which consists of two phases. In the offline mining phase,the net profit objective function is used to evaluate the potential profit of the driving route,and the high-profit road segment area is clustered by DBSCAN. In the online recom-mendation stage,taxi drivers are provided with high-efficiency routes based on the current location of the taxi combined with the actual road conditions feedback from the current section. The framework is especially sensitive to special situations (weather,performances and other large events),and it can make multiple recommendations for popular routes. The experiment uses Haikou’s onemonth taxi GPS recorded data. The final experimental results verify the correctness and effectiveness of the PTRA.