Mining Taxis' Trajectories For Cruising Routes Recommendation

As it becomes more and more convenient to capture moving objects' trajectories, a lot of location based services(LBS) are devised based on the big data. These services has important theoretical significance and practical values for citizens' life, road planning and urban development.Taxis are playing an essential role of transportation system, which are cruising in every corner of the city. Its trajectories are generally public without privacy issues, together with the high precision and the strong continuity of positioning. Therefore, the taxis' trajectories become popular on industrial applications and academic researches.This paper focused on mining taxis' trajectories for cruising routes recommendation. Aiming to find taxi-passengers as soon as possible, some deep researches are developed, such as mining trajectories' spatial-temporal features, drivers' preference analysis of position types or cruising routes, and the effects which manifold attributes of pickup spots exert on routes recommendation. This article will do benefit to both increase drivers' profits and reduce the pollution of the fuel consumption.The main contribution of the paper lies in:(1) A novel approach of spatial temporal analysis(STA) on pickup points is proposed based on the density clustering algorithm. The passenger-finding locations are mined by analyzing the spatial-temporal distribution characteristics of pickup points. Experimental results show that STA can improve the accuracy of passenger-finding.(2) A personalized recommendation method for finding passengers is put forward based on the locations' contents. The experiments show that taxi-drivers usually select passenger finding locations according to their preferences.(3) Information entropy based cruising routes recommendation is brought forward. We selected four factors from multiple factors that affect the taxi driver to travel to a passenger finding location, to measure the weight they affect the driver to select location. The experiments' result shows that our method is excellent. The performance in recommendation for no-load taxi can get more benefits.