利用社区选举和链路预测的分类方法  

Classification Method Using Community Election and Link Prediction

作　　者：杨旭华[1] 陈孝则 王磊[1] 许营坤 叶蕾[1] 毛剑飞[1] YANG Xu-hua;CHEN Xiao-ze;WANG Lei;XU Ying-kun;YE Lei;MAO Jian-fei(College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学计算机科学与技术学院,杭州310023

出　　处：《小型微型计算机系统》2022年第8期1569-1576,共8页Journal of Chinese Computer Systems

基　　金：国家自然科学基金面上项目(61773348)资助;浙江省自然科学基金项目(LY20F020029)资助.

摘　　要：分类方法通过比较数据之间的相似性,把不同特征或属性的数据分别归属到不同的类别,在金融、医学和生物等领域有着广泛的应用.本文首次提出了一种利用社区选举和链路预测的分类方法(CCELP),该方法首先用k近邻算法将数据集转化成一个稀疏网络,接着使用社区检测算法把网络划分为多个社区,并通过投票选举得到每个社区的代表节点,移除不符合“过半数原则”的部分代表节点,将剩余代表节点同社区内节点相连得到新网络;接着提出了考虑二级共同邻居的分类链路预测(CLP)指标,在新网络中按照节点和代表节点间的CLP指标把节点归属到不同的类别中去,从而完成数据分类.在16个数据集上,CCELP与8种知名分类方法进行了比较,实验结果表明CCELP具有优异的分类效果.The classification method compares the similarity between data and assigns data with different characteristics or attributes to different categories.It has a wide range of applications in the fields of finance,medicine and biology.In this article,a classification method using community election and link prediction(CCELP)is proposed for the first time.The method first uses the k-nearest neighbor algorithm to transform the dataset into a sparse network,and then uses the community detection algorithm to divide the network into multiple community,and get the representative nodes of each community through voting,remove some representative nodes that do not conform to the“majority rule”,and connect the remaining representative nodes with nodes in the community to obtain a new network;then proposes the consideration of second-level common neighbors classification link prediction(CLP)index,in the new network according to the CLP index between the node and the representative node,the nodes are assigned to different categories,thereby completing the data classification.On 16 datasets,CCELP is compared with 8 well-known classification methods.The experimental results show that CCELP has excellent classification effects.

关 键 词：社区检测 链路预测 机器学习 社区选举 分类算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]