Abstract: This paper proposes a novel approach for continuous approximate query over trajectory stream based on sliding window. Through local clustering, the sliding window is divide into various sized basic windows and sampling the data elements of a basic window using biased sampling rate, forms trajectory stream synopses. Toward such synopses, it can implement distributed K-Nearest Neighbor(KNN) queries utilizing the plane sweep algorithm of computational geometry. The extensive experiments verify the effectiveness of proposed algorithm and it has better expansibility.

Key words: trajectory data stream, local clustering, biased sampling, data synopses, K-Nearest Neighbor(KNN) query

摘要： 提出一种基于滑动窗口的K-最近邻(KNN)近似查询算法。将滑动窗口内数据通过聚类划分成若干大小不一的基本窗口，针对每个基本窗口给定一个采样率，对窗口内数据进行偏倚采样，形成数据流摘要，并基于该摘要，采用计算几何平面扫描算法执行分布式最近邻查询。仿真实验结果表明该算法有效，且具有较好的可扩展性。

关键词: 轨迹数据流, 局部聚类, 偏倚采样, 数据摘要, K-最近邻查询

CLC Number: 

• TP311