摘要
目的提高算法效率,减少磁盘访问次数,提出一种基于密度的高效增量聚类算法ISNN.方法将更新对象的空间进行划分,定义了基于该划分的最近邻居概念,在此基础上应用一种剪枝策略来确定受影响对象的集合,数据更新时,只需要对受影响对象集合进行处理.结果受影响对象集合远小于原数据集合,显著地提高了算法效率.结论实验表明,ISNN在效率和磁盘访问次数上都显著优于SNN算法.
This paper proposes an incremental algorithm, ISNN which is based on density-based clustering algorithm SNN. The algorithm partitions the space around the update object, and redefines the nearest neighbors in each partition. In addition, a prune strategy is adopted; in this way we can find the influenced object dataset. When updating, the algorithm only deals with the set of the influenced objects instead of the whole dataset. Since the size of the influenced object dataset is far smaller than that of the whole dataset, the performance of the algorithm is improved. The evaluation shows that ISNN has much better efficiency and less I/O processing than SNN.
出处
《沈阳建筑大学学报(自然科学版)》
CAS
2006年第6期1015-1018,共4页
Journal of Shenyang Jianzhu University:Natural Science
基金
辽宁省自然科学基金(20052006)
辽宁省教育厅攻关计划(05L354)