随着标签分类应用的增长,社交网络环境下多标签分类已成为一个重要的数据挖掘研究领域。关系分类模型基于一阶邻居做标签分类,其性能优于传统的多标签分类器。但现有的关系分类模型也存在问题:第一,仅利用一阶邻居做分类,未能充分使用邻居信息。第二,网络数据通常包含大量不连通的孤立部分,其标签无法利用现有的关系分类模型分类。考虑基于共引规则为非孤立节点挖掘二阶邻居和基于节点特征向量相似度为孤立节点挖掘高阶邻居,本文提出一种新的基于多阶邻居的网络数据多标签分类算法,称为MORN算法。在多个真实数据集上将MORN与现有的关系分类模型作对比,实验表明,MORN算法能够学习到更多节点的标签且精度优于传统关系分类方法。%Multi-label classification in network environments is becoming a key area of data mining research as its ap-plications are increasing dramatically.Relational classification models,which predict class labels of linked neighbors accord-ing to the ones of the given nodes,have been shown to outperform traditional multi-label classifiers.However,existing rela-tional classification models neither make full use of neighbor information,nor predict the isolated nodes’labels,which are popularly existing in relational networks.In this paper,we present a multi-label relational classifier (MORN)that mines both second-order neighbors for non-isolated nodes and high-order neighbors for isolated nodes.MORN has been conducted on real datasets and it demonstrates that our proposed classifier outperforms existing relational classification models.
展开▼