基于稀疏子集分析的轨迹聚类发现

Track Clustering Discovery Based on Sparse Subset Analysis

论文中分析了城市出租车的轨迹和检测社会集群。为了更好地检测轨迹中可能的群落,提出了一种鲁棒的社区检测算法。该算法从轨迹集中提取数据特征空间关系矩阵,并使用这个矩阵定义轨迹相似度矩阵。将相似度矩阵变换为相异度矩阵,基于相异度的稀疏子集选择(DS3)算法用于分析多个稀疏子集。每个子集对应于一个集群,该集群是要检测的社区。这可以避免陷入局部最优,不需要进行算法迭代和多次计算来提高社区检测的准确性和效率。实验结果证明了该方法的有效性。

In this paper,the trajectories of urban taxis is analyzed and social communities are detected.In order to better de⁃tect the possible communities in the trajectory,a robust community detection algorithm is proposed.This algorithm firstly extracts one feature matrix,i.e.,the spatial relation matrix,from the trajectory set,which is employed to define the trajectory similarity ma⁃trix.Then the similarity matrix is transformed into a dissimilarity matrix,and sparse subset selection(DS3)algorithm based on dis⁃similarity is used to analyze multiple sparse subsets.Each subset corresponds to a cluster,which is the community to be detected.This can avoid falling into local optimum,there is no need to perform algorithm iteration and multiple calculations to improve the ac⁃curacy and efficiency of community detection.Experimental results demonstrate the effectiveness and efficiency of this method.