Discovering top-weighted k-truss communities in large graphs

Abstract

Abstract Community Search is the problem of querying networks in order to discover dense subgraphs-communities-that satisfy given query parameters. Most community search models consider link structure and ignore link weight while answering the required queries. Given the importance of link weight in different networks, this paper considers both link structure and link weight to discover top-r weighted k-truss communities via community search. The top-weighted k-truss communities are those communities with the highest weight and the highest cohesiveness within the network. All recent studies that considered link weight discover top-weighted communities via global search and index-based search techniques. In this paper three different algorithms are proposed to scale-up the existing approaches of weighted community search via local search. The performance evaluation shows that the proposed algorithms significantly outperform the existing state-of-the-art algorithms over different datasets in terms of search time by several orders of magnitude.

Keywords

• Community search
• Weighted graph
• k-truss community detection model