フシミ　タカヤス   Takayasu Fushimi   伏見　卓恭    所属   コンピュータサイエンス学部 コンピュータサイエンス学科    職種   専任講師
言語種別	日本語
発行・発表の年月	2017/06
形態種別	学術論文
査読	査読あり
標題	メトリック空間における複数カテゴリに属するハイブリッドオブジェクト抽出法の提案
執筆形態	共著
掲載誌名	情報処理学会論文誌 ネットワーク科学特集号
掲載区分	国内
著者・共著者	伏見 卓恭，斉藤 和巳，風間 一洋
概要	In this paper, we propose two centrality measures, mixedness and cohesiveness centrality, intended to extract representative objects in a metric space. These measures are based on betweenness and closeness centrality each of which is widely used in the social network analysis field in order to extract important nodes. The mixedness centrality is an indicator that quantifies the degree to be located medway between any pairs of objects and extracts mixed objects located in medway of clusters. By contrast, the cohesiveness centrality is an indicator that quantifies the degree to be located in densely distribution and extracts cohesive objects located in center of densely distributed objects. In our experiments using several types of real and synthetic datasets, we show that our proposed centrality measures can extract characteristic representative objects in each dataset.