基于改进度中心性的样本点相似性度量方法  

Improvement centrality-based Similarity measure method for patterns

作　　者：邓莉 刘士虎[1] DENG Li;LIU Shi-hu(School of Mathematics and Computers,Yunnan Minzu University,Kunming 650504,China)

机构地区：[1]云南民族大学数学与计算机科学学院,昆明650504

出　　处：《云南民族大学学报（自然科学版）》2022年第4期425-432,共8页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：国家自然科学基金(61966039);云南省教育厅科学研究基金(2021Y670).

摘　　要：针对带权图数据中样本点的相似性度量问题,从图数据中样本点的属性信息和拓扑信息2个角度出发,提出了一种基于改进度中心性的样本点相似性度量方法.首先,该方法在充分研究了任意2个样本点拓扑信息可达的基础上,以改进度中心性指导随机游走的方法,度量样本点基于拓扑信息的相似性.其次,考虑到传统欧式距离在度量相似性方面的不足,提出一种融合的度量方法来度量样本点基于属性信息的相似性,进而得出样本点间的综合相似性.最后,通过一个具体的算例来验证所提出方法的实用性和有效性.This paper discusses the similarity of patterns in weighted graph data.Here,the so-named weighted graph data compose two absolutely different information,one is pattern's attribute information and the other is the relation between patterns,i.e.,topological information.Bearing this in mind,an improvement centrality-based similarity measure method for patterns is proposed in this paper,abbreviated as ICSMMP.In this measurement,the reachability of any two patterns is fully studied,and then,the similarity of topological information is measured by random walk that guided by improved degree centrality.Meanwhile,considering the shortcomings of Euclidean distance in measuring similarity based on attribute information,we propose an integrated method to measure the distance of patterns.On this base,a comprehensive and novel similarity measure method for patterns in graph data is constructed.In conclusion,a concrete example is given to illustrate the practicability and effectiveness of the proposed measurement.

关 键 词：带权图数据 样本点 改进度中心性 随机游走 相似性度量 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]