作　　者： ; ;

机构地区： 暨南大学信息科学技术学院计算机科学系

出　　处： 《计算机工程与设计》 2006年第11期1939-1942,共4页

摘　　要： 增量式挖掘方法有适应大规模动态数据、降低内存需求和可实现并行处理等诸多好处,但是目前的增量式聚类方法存在参数限制较多和计算结果不够准确等问题。在信息源变化的数据挖掘体系结构下,利用一群特殊的智能代理增量修改知识模型,提出了群体智能聚类模型的构建方法及增量模型维护算法。该方法利用信息熵加快聚类过程,根据信息素和数据库的插入及删除增量操作调整已生成的聚群,设定的参数较少,实验表明聚类结果准确。 Incremental data mining is to adopt large-scale dynamic data, to reduce storage requirement and to implement parallel processing, but presented incremental clustering methods exist such the problems as there are comparatively many parameters and the results is not exact enough. In the data mining structure with continuous changing sources, using a swarm of special Agents to modify knowledge model incrementally, a clustering model based on swarm intelligence and an incremental maintenance algorithm are presented. The information entropy is applied to speed up clustering and modifies existing clusters according to pheromone and incremental insertions and deletions of database, which needs a few parameters. Experiments prove that it could get almost exact clusters.

关 键 词： 增量 数据挖掘 群体智能 聚类 信息熵

领　　域： [自动化与计算机技术] [自动化与计算机技术]