基于快搜索和寻找密度峰值的聚类探究

rajectory data records the track information of time and space. Nowadays huge amount of trajectory data is collected and researched. By clustering, people could find out similar groups moving objects easily, which is always considered as the first stage for trajectory data mining. Because the data form of trajectory is far from the traditional data type, trajectory clustering always requires massive computing time and space. For most trajectory distance measures, O(nm) complexity are required, where n and m represent the lengths of two trajectories. Moreover, conventional clustering methods, like K-Means, K-Medoids, need iterating until converging, which also take lots of time. This work utilizes Douglas Peucker algorithm to simplify trajectories and computes the Dynamic Time Warping distance, then investigates a new clustering method proposed by Rodriguez and Laio (2014), fast search and find of density peaks, working on the trajectory clustering.