异质信息网络快速离群点数据智能挖掘算法  

Intelligent Mining Algorithm for Fast Outlier Datain Heterogeneous Information Networks

作　　者：庄巧蕙 Zhuang Qiaohui(School of Economics and Management,Minnan University of Science and Technology,Shishi,Fujian 362700,China)

机构地区：[1]闽南理工学院经济管理学院,福建石狮362700

出　　处：《黑龙江工业学院学报(综合版)》2024年第10期97-101,共5页Journal of Heilongjiang University of Technology(Comprehensive Edition)

基　　金：福建省中青年教师教育科研项目(项目编号:JAT220426);闽南理工学院校级科研项目“双层网络中心非显著特征数据挖掘的算法”(项目编号:19KJX013)。

摘　　要：为检测出异质信息网络中潜在的异常行为,提升数据质量和决策准确性,研究异质信息网络快速离群点数据智能挖掘算法。分析异质信息网络中离群点的成因,在MapReduce模型基础上搭建分布式集群,在集群子节点中应用网格算法初步获取候选离群点集,并将这一候选离群点集发送至分布式集群中的主节点上,在主节点应用基于密度的离群点挖掘算法,计算候选离群点数据集中每一对象的局部离群点因子,通过局部离群点因子排序,挖掘出异质信息网络的离群点数据。经实验验证:该算法可保持精准的离群点数据挖掘效果,针对不同数据集挖掘时具有较高的AUC接收者操作特征以及较低的误报率,具有良好的应用效果。To detect potential abnormal behaviors in heterogeneous information networks and improve data quality and decision-making accuracy,an intelligent data mining algorithm for fast outliers in heterogeneous information networks was studied.Analyze the causes of outliers in heterogeneous information networks,build a distributed cluster based on the MapReduce model,apply a grid algorithm to obtain a preliminary set of candidate outliers in the sub nodes of the cluster,and send this set of candidate outliers to the main node in the distributed cluster.On the main node,apply a density based outlier mining algorithm to calculate the local outlier factor of each object in the candidate outlier dataset,by sorting local outlier factors,outlier data of heterogeneous information networks can be mined.Experimental verification shows that this algorithm can maintain accurate outlier data mining performance,with high AUC receiver operation characteristics and low false alarm rates for mining different datasets,and it has good application effects.

关 键 词：异质信息网络 离群点数据 智能挖掘 分布式集群 网格算法 离群点因子 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]