Recommend Taxi Pick-up Hotspots Based on Density-based Clustering

Abstract

Mining valuable information from taxi trip data to recommend taxi drivers hotspot pickup areas become a hot research problem since taxis cruising in the city causes large energy waste every day. In existing methods, many methods are just cluster pick-up hotspot areas by various clustering algorithm without further analysis of the differences between the hotspots. In this paper, we propose a novel taxi pick-up recommendation model analyzing hotspot areas according to different factors based on an improved DBSCAN algorithm. We conduct several experiments with synthetic datasets and a real taxi trip dataset to illustrate cluster results and a real taxi trip data to verify the efficiency of algorithm and the precision of the recommendation model. The experiment results show that the proposed algorithm is capable for efficiently and effectively detecting clusters with multiple density-levels automatically with different multiple density-levels and the proposed taxi pick-up hotspots recommendation model has higher precision than other methods.