基于查询集空间分布的聚合最近邻查询算法

摘要

Aggregate nearest neighbor query involves many query points, so it is more complicated than traditional nearest neighbor query, and the distribution characteristic of query set implies the region where its aggregate nearest neighbor exists. Taking full account of the distribution characteristic of query set, a method by utilizing distribution characteristic to direct the way of aggregate nearest neighbor searching was given. Based on the method, a new algorithm named AM was presented for aggregate nearest neighbor query. AM algorithm can dynamically capture and use the distribution characteristic of query set, which enables it to search data points in a right order, and avoid unnecessary searching to data points. The experimental results show the efficiency of the algorithm.%聚合最近邻查询涉及到多个查询对象,因此比传统最近邻查询更复杂,而且其查询集空间分布特征暗含了查询集聚合最近邻的区域分布信息.充分考虑查询集分布特征,给出了利用分布特征指导聚合最近邻搜索的方法,并以此提出了一种新的聚合最近邻查询算法——AM算法.AM算法能动态地捕捉并利用查询集空间分布特征,使得对数据点的搜索按正确的次序进行,避免对不必要数据点的搜索.最后通过实验验证了AM算法的高效性.