CAJ | 학술논문

针对社交网络的有向交互性和大规模特性，该文提出一种基于结构相似度的有向网络聚类算法(DirSCAN)，以及相应的分布式并行算法(PDirSCAN)。考虑社交网络中节点间的有向交互性，将行为结构相似的节点聚集起来，并进行节点功能分析。针对社交网络规模巨大的特点，提出MapReduce框架下的分布式并行聚类算法，在确保聚类结果一致的前提下，提高处理性能。大量真实数据集上的实验结果表明，DirSCAN比无向网络聚类算法(SCAN)在F1上可提高2.34%的性能，并行算法PDirSCAN比DirSCAN运行速度提升1.67倍，能够有效处理大规模的有向网络聚类问题。
침대사교망락적유향교호성화대규모특성，해문제출일충기우결구상사도적유향망락취류산법(DirSCAN)，이급상응적분포식병행산법(PDirSCAN)。고필사교망락중절점간적유향교호성，장행위결구상사적절점취집기래，병진행절점공능분석。침대사교망락규모거대적특점，제출MapReduce광가하적분포식병행취류산법，재학보취류결과일치적전제하，제고처이성능。대량진실수거집상적실험결과표명，DirSCAN비무향망락취류산법(SCAN)재F1상가제고2.34%적성능，병행산법PDirSCAN비DirSCAN운행속도제승1.67배，능구유효처리대규모적유향망락취류문제。
To cluster the directed and large-scale social networks, a Structural Clustering Algorithm for Directed Networks (DirSCAN) and a corresponding Parallel algorithm (PDirSCAN) are proposed. Considering oriented behavioral relation between two vertices, DirSCAN is constructed based on action structural similarity and function analysis. To meet the need of large-scale social network analysis, a lossless PDirSCAN based on MapReduce distributed parallel architecture is designed to improve the processing performance. A large number of experimental results on real-world network datasets show that DirSCAN improves performance of SCAN up to 2.34% on F1, PDirSCAN runs 1.67 times faster than DirSCAN.