基于聚类分析局部离群点挖掘改进算法的研究与实现  

ON IMPROVED ALGORITHM FOR LOCAL OUTLIER MINING BASED ON CLUSTER ANALYSIS AND ITS IMPLEMENTATION

作　　者：赵战营[1] 成长生[1] 

机构地区：[1]东华大学计算机科学与技术学院,上海201620

出　　处：《计算机应用与软件》2010年第11期255-258,共4页Computer Applications and Software

摘　　要：对于犯罪检测、网络入侵检测等应用,离群点检测是数据挖掘的一种重要算法。局部离群因子是对数据对象离群点的程度定义,计算所有数据对象局部离群因子需要大量计算。一种基于聚类分析局部离群点挖掘改进算法得以实现,此改进算法以聚类分析为预处理,只对聚类之外的数据对象计算局部离群因子,避免了大量计算,并改进了对数据对象k距离邻域的求解。通过仿真数据和轨道交通AFC(automatic fare collecting system)客流数据的实验,证实此改进算法不仅能更高效地挖掘出值得关注的离群点,而且还能更好地达到解析目的。Outlier detection is an important algorithm in data mining for applications such as criminal activities＇ detecting and network intrusion detection,etc.Local outlier factor is a grade definition for outlier of data object,a great deal calculation is required to calculate all the local outliers of data objects.An improved algorithm for local outlier mining based on cluster analysis is implemented in this paper,in which the cluster analysis is taken as a preprocessing,local outlier factors are calculated only for the data objects out of cluster,so a great deal calculation is avoided and k-distance neighbours searches of data object are improved.It is proved from the simulating data and the experiments of passenger flow dataset of automatic fare collecting system of rail transit that the improved algorithm can mine out the outliers deserving attention more effectively,and can also achieve the parsing aims better.

关 键 词：数据挖掘 局部离群因子 K-距离邻域 聚类分析 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]