数据仓库中基于密度的批量增量聚类算法

数据仓库为数据挖掘提供了很好的平台,当数据仓库中的数据发生变化时,原来挖掘出来的模式也要相应地进行更新。MartinEster等最先提出了增量聚类算法,但算法在增量聚类过程中,更新对象依次一个个地单独处理,而没有考虑更新对象之间的关系,效率较低。该文提出了基于DBSCAN算法的批量增量聚类算法,减少了对象的检索,提高了增量聚类的效率。

An Incremental Density-based Clustering Algorithm in a Batch Mode Used in a Data Warehouse

Data warehouses provide a great deal of opportunities for performing data mining.Typically,updates are collected and applied to the data warehouse in a batch mode.Then,all patterns derived from the warehouse have to be updated as well.Martin Ester introduced the first incremental clustering algorithm.But in the algorithm,sets of updates are processed one at a time without considering the relations between the single updates.In this paper,we present a incremental clustering algorithm based on DBSCAN in a batch mode,which improves the efficiency of pattern updates greatly by reducing the retrieve of updated objects.