The recommendation algorithm for taxi drivers based on hadoop and historical trajectory of taxis

In order to solve the problem of "Hard to take a taxi, hard to find a passenger" in cities, we present a recommendation algorithm for taxi drivers based on historical trajectories of taxis and mathematical statistics. We offer taxi drivers the path with the minimum time expectation composed of the cruising time on the path and waiting time at parking places. Specifically, we make use of Hadoop platform to deal with historical trajectories and we also improve a clustering algorithm and implement it in parallel on Hadoop. To quantify people's tolerance to the weather precisely, we introduce the human body comfort meteorological index; meanwhile, the S-curve is used to fit the relation between weather conditions and the taxi demand. The experiments preformed in the field show that the accuracy of the recommendation can reach 87.5%, which demonstrates the effectiveness of the algorithm.