Recommender Systems For Taxi Drivers With Spatio-temporal Constraints

Recent years,with the proliferation of embedded sensors and short-range wireless communication technologies,intelligent transportation services have emerged as a useful paradigm to improve transportation systems.Among these intelligent services,the taxi recommender system that provides real-time suggestions to the drivers is an essential part.With the aid of vehicle sensor data and Global Positioning System(GPS)traces,the taxi recommender systems can significantly benefit the entire taxi business by avoiding traffic jams and reducing passengers’waiting time,drivers’ cruising time and even fuel consumption.In reality,taxi drivers may have their preferred destination regions or target travelling time to avoid traffic jams or to execute arranged pickup orders.However,no prior work managed to maximize the profit of drivers and satisfy the requirements for destination regions and travelling time at the same time.To tackle these challenges,we propose two recommendation system under these two constraints.The main contents of this article include the following three points:(1)Consider the factors that affect the recommendation of taxi search routes.By mining the taxi GPS trajectory data,we use the HDBSCAN algorithm to recommend the potential pick-up points and points of interests to taxi drivers.Then we match the real pick-up points with the extracted area(2)Consider limiting the destinations of passengers under the destination constraint.We recommend a suitable route for the driver.Through the clustering analysis of the potential pick-up point area,we get the points of interests.Then we calculate the inter-regional probabilities and establish a route recommendation algorithm model to maximize the potential profit of taxi drivers(3)Consider limiting the length of the passenger’s order journey under the time constraint.We recommend a route for the taxi drivers.Through statistics on the length of historical passenger travelling time,we calculate the point-to-time probabilities and establish a route recommendation algorithm model to maximize the potential profit of taxi driversThe experimental results based on the taxi traces collected from Shanghai,China,validate the effectiveness of the proposed recommender system.