摘要
聚类是一种智能算法,该技术在数据挖掘中的一个重要分支技术。在现有的各种聚类算法中,密度聚类算法有着广泛的应用。密度算法具有与其他聚类算法聚类时的不同之处是密度聚类是否聚类成功决定于参数Eps和Min Pts的取值。因此目前对于密度聚类算法的研究仍然是基于对聚类参数Eps和Min Pts的研究。论文参考了MMTD算法和粗糙集中的决策系统之后,采用这两种算法对密度聚类法中的Min Pts参数进行研究。论文对DBSCAN算法的研究思路是:当进行密度聚类时首先使用MMTD算法对参数Min Pts时进行估量,其次再使用决策系统对MMTD做出的估量做出决策判断,最后根据决策判断的结果来判断这次聚类是否成功。
CIustering is an inteI igent aIgorithm, which is an important branch of data mining. In the existing cIustering aIgorithms, the density cIustering aIgorithm has a wide range of appIications. The density aIgorithm has the dif erence with other cIustering aIgorithms, which is based on the cIustering of the density and whether the cIustering is determined by the vaIues of Eps and MinPts. So far, the research on density cIustering aIgorithm is stiI based on the cIustering parameters Eps and MinPts. After the MMTD aIgorithm and the decision system of rough set, the two aIgorithms are used to study the MinPts parameters in the density cIustering method. In this paper, we study the DBSCAN aIgorithm is:when the density of the first use of MMTD aIgorithm to estimate the parameters of the MinPts, and then use the decision-making system to make the decision to make the decision of the MMTD, the finaI decision to determine whether the cIustering is successfuI.
基金
北京航空航天大学软件开发环境国家重点实验室开放基金资助项目(SKLSDE-2013KF-02)
