Efficiently Conditional Community Search in Dynamic Graphs

Abstract

In recent years, community search have been extensively studied in a variety of practical applications, including friend recommendation, event organization, protein complex identification, and so on. Community search is to find a densely connected sub graph based on query nodes in a graph. In reality, however, users may also have opinions or requirements about other nodes in the graph other than the query nodes. In this situation, a new type of community search has emerged under complex user search conditions, which is called conditional community search. The current definitions and solutions of conditional community, however, are not very applicable since the user search conditions have become more and more complex. Moreover, real graphs often involve as the time goes on. To address this issue, therefore, we investigate the problem of conditional community search in dynamic graphs. To the best of our knowledge, this is the first work for finding conditional communities from dynamic graphs. First, we present a new definition of conditional community with regard to multiple user search conditions. Then, we propose an effective search solution for finding conditional communities in static graphs, and this solution can be used as a basic method to search conditional communities in dynamic graphs. And then we explore a novel incremental search solution for searching conditional communities in dynamic graphs. Finally, we conduct comprehensive experimental studies on several real datasets to evaluate the effectiveness and the efficiency of our solutions. The experimental results show the accuracy and efficiency of our incremental search solution for the problem of conditional community search in dynamic graphs.