【作者】 张珂；

【Author】 Zhang Ke;Department of Civil Engineering,Tsinghua University;

【机构】 清华大学土木工程系；

【摘要】 城市交通拥堵会影响居民的日常出行与城市运行效率,因此挖掘分析城市交通拥堵的规律性以及通过聚类方法研究居民出行的相似性,有助于交通管理部门更好地处理"城市病"难题,同时也会为更舒适的出行路径选择提供参考。传统的聚类算法容易受到异常轨迹数据的影响,本文根据北京市奥林匹克公园附近的车辆轨迹信息,研究了交通拥堵规律与POI,时间,空间等因素的相关性影响,然后改进了ST-DBSCAN算法来融合方向角、瞬时速度的信息。在对比实验中,本文选取DTW作为评价指标,并发现融入方向角信息可以有效提升聚类结果,而瞬时速度会对聚类效果产生负面影响。参数的敏感性分析进一步探索了阈值设定对聚类结果的影响程度。更多还原

【Abstract】 Urban traffic congestion affects residents’ daily travel and urban operation efficiency.Therefore,analyzing the regularity of urban traffic congestion and studying the similarity of residents’ travel through clustering method can help traffic management department to better deal with urban disease,and also provide reference for the choice of travel routes.Traditional clustering algorithms are likely to be affected by abnormal trajectory data.According to the vehicle trajectory information near Beijing Olympic Park,this paper studies the correlation between traffic congestion and POI,temporal,space and other factors,and then improves the ST-DBSCAN algorithm to fuse the information of direction angle and instantaneous speed.In the comparative experiment,this paper selects DTW as the evaluation indexes,and concludes that integrating the direction angle information can effectively improve the clustering results,while the instantaneous speed has a negative impact on the clustering effect.Parameter sensitivity analysis further explored the influence of factors on the clustering results.更多还原