Finding top-k influential set in directed graphs

Abstract

A RkNN (Reverse k Nearest Neighbor) set of a query q is known as the influence set of q that contains the top k influential points. This query has got a considerable amount of attention due to its importance in various applications involving decision support system, profile based marketing, location based services, etc. Although this query has been largely studied in Euclidean spaces but there is very less work done in the context of large graphs. In this dissertation, a framework has been proposed for RkNN query over directed graphs. We present a heuristic that cuts out the search space substantially for finding out RkNN of a query point. To the best of our knowledge there has been no work done for RkNN query over directed graphs. We conduct extensive experiments over some real world data sets like DBLP, social network, product co-purchasing network of Amazon and study the performance of the proposed heuristic in various settings. Experiments show the effectiveness of the proposed heuristic. We also study a co-authorship application in the context of RkNN query over undirected graph, wherein we design a metric to define the similarity between two authors. Here RkNN query can be interpreted as a query to find out the influence set of an author from an authors collaboration network (DBLP).