基于重心点转移的St-DBSCAN改进算法

An Improved St-DBSCAN Algorithm Based on Center of Gravity Shifting

在目前已提出的聚类算法中,St-DBSCAN算法是一种基于密度且性能优越的时空聚类算法。但是当时空点分布出现密度倾斜时,St-DBSCAN算法会出现聚类时间过长和聚类效果不好的问题。基于此,通过对空间点分布存在的三种数据倾斜,采用数据重心点转移策略,提出了对应的解决方案,以此实现了改进后的St-DBSCAN算法。为了验证改进后算法的性能,以昆明市出租车GPS数据为实验数据,进行了算法性能对比实验。实验结果表明,改进St-DBSCAN算法的时间性能和聚类效果有了一定程度的提升。

In the presented clustering algorithms,the St-DBSCAN is a spatio-temporal clustering algorithm based on density with superior performance.However,when the spatial distribution is tilted,the St-DBSCAN algorithm may produce too long clustering time and poor clustering effect.Based on the problem,we propose the corresponding solution by using the data center point transfer strategy for three kinds of data skew in spatial point distribution,and then implement the improved St-DBSCAN algorithm.In order to verify the proposed algorithm,the GPS data of taxi of Kunming is used as experimental data for performance comparison,which shows that the improved St-DBSCAN algorithm is improved in time performance and clustering effect.