考虑需求的出租车收益路线推荐

Profitable taxi route recommendation considering demand

当前研究忽略出租车接载下一个乘客获得的收益,并且会为出租车推荐具有较低接客概率的路线.针对以上问题,基于移动顺序推荐对出租车路线推荐问题进行建模,提出考虑出租车下一个乘客收益的出租车收益路线推荐(PMSR)的改进模型.考虑到上客点的需求大小对出租车在上客点成功接载乘客的可能性的影响,提出考虑需求的出租车收益路线推荐(PMSR-D)的改进模型.基于上海市出租车GPS轨迹数据,采用带噪声的基于密度的空间聚类(DBSCAN)算法、模拟退火算法和贪心算法对PMSR和PMSR-D模型进行验证.结果表明,PMSR模型推荐路线的上客点的最低预期票价较高,其接客概率与潜在巡航距离(PTD)、基于时空矩阵的路线推荐(RTS)模型相比在7:00—10:00和13:00—16:00分别平均增加了148.2%、253.0%和88.1%、48.0%,表明PMSR模型能够为出租车推荐预期票价较高且接客概率更大的路线,优于PTD和RTS模型.与PMSR模型相比,PMSR-D模型在7:00—10:00和13:00—16:00分别增加了125和20个潜在的乘客需求,验证了PMSR-D模型的有效性.

The existing research neglects the income earned by taxi driver when picking up the next passenger and recommends route with a relatively low probability of picking up for taxi.In response to the above problem,the taxi route recommendation problem based on mobile sequential recommendation was modeled,and an improved model for profitable taxi route recommendation based on mobile sequential recommendation(PMSR)considering the next passenger’s income was proposed.An improved model for profitable taxi route recommendation based on mobile sequential recommendation considering demand(PMSR-D)was proposed,considering the impact of the demand for pick-up point on the likelihood of taxi successfully picking up passenger at pick-up point.The density-based spatial clustering of applications with noise(DBSCAN)algorithm,simulated annealing algorithm and greedy algorithm were used to verify the PMSR and PMSR-D models,based on the taxi GPS trajectory data in Shanghai.Results showed that the minimum expected fare at the pick-up points recommended by PMSR model was relatively high.From 7:00 to 10:00,the PMSR model had an average increase of 148.2%and 253.0%in picking up probability compared to the potential cruising distance(PTD)model and the route recommendation model based on temporalspatial metric(RTS),respectively.From 13:00 to 16:00,the PMSR model had an average increase of 88.1%and 48.0%in picking up probability compared to the PTD and RTS model,respectively.This indicated that the PMSR model can recommend route with high expected fare and high picking up probability for taxi,which was superior to the PTD and RTS models.Compared with the PMSR model,the PMSR-D model added 125 and 20 potential passenger demands from 7:00 to 10:00 and 13:00 to 16:00,respectively,verifying the effectiveness of PMSR-D model.