Mining taxi trajectories for most suitable stations of sharing bikes to ease traffic congestion

Sharing bike service is a new emerging public transportation, which has been the hottest topic for months. Sharing bike service provides flexible demand‐oriented transit services for city commuters. However, as large amount of sharing bikes flood into big cities, problems caused by chaotic order of sharing bikes are emerging slowly. The authors aim to draw support from taxi trajectories to analyse current traffic condition and improve it with sharing bikes. In this study, the authors propose a traffic congestion finding framework, called CF. In CF, derived from the points density‐based clustering method of inspiration, the authors propose a new clustering method (CF‐Dbscan) and successfully applied it to the clustering of trajectories. A road network matching algorithm (CF‐Matching) helps to match GPS points to road net even if points are in low‐sampling‐rate. They also employ a ranking feedback mathematical model to adjust the number of sharing bikes of different stations to meet people's demand and reduce redundancy. The first experiment proves that the proposed clustering algorithm performs better than traditional DBSCAN. Another experiment is conducted to verify the effectiveness of the proposed framework in reducing traffic congestion. The experimental results prove that with the proposed framework the authors can achieve the purpose of easing traffic congestion.