关于出租车载客地点序列推荐技术的研究

在大城市中，出租车已成为实现智能交通运输系统不可或缺的一环。然而，由于一些出租车司机的驾驶经验，和对城市活动的熟悉程度的不足，使得其在寻找乘客时会采取毫无目的的随机漫游策略。这就导致了出租车司机的收益不高，同时也造成了能源的消耗以及环境的污染。针对此问题，将提出出租车载客地点的推荐模型，使得模型给出的推荐地点序列能获得较高的期望收益。具体来说，将基于出租车GPS轨迹数据建立出租车载客地点的马尔科夫决策过程模型，并给出求解该模型的2种算法。仿真实验结果显示，与典型的TopK方法相比，给出的推荐结果能更好地提高单位时间内出租车司机的收益。

Towards Sequential Recommendation of Passenger- finding Locationsin An Urban Taxi Service

In modem cities, taxis play a quite significant role in intelligent transportation system. However, due to lacking of enough driving experience and knowledge of cities, some taxi drivers tend to take stochastic cruise for finding passengers when they are in vacant. This leads to the low profit as well as the energy consumption and environment pollution. This pa- per presents a recommender model for taxi driver to recommend a series of locations in which taxi driver can get high expec- ted profit for passenger finding. Specifically, the paper establishes a Markov Decision Process model which is based on taxi- cab＇s GPS trajectory data, and two algorithms for solving this model will be given. Simulation results show that this model could gain better recommendation performance than TopK method, with the metric of profit per unit time.