Development of multidimensional academic information networks with a novel data cube based modeling method

A common task in many applications is to find people who are knowledgeable about a given topic, topics which are suitable for a given author or venue, and venues which are attractive for a given author or topic. This problem has many real-world applications and has recently attracted considerable attention. However, the existing methods are not very efficient in providing flexibility for multi-dimensional and multi-level view from different perspectives. In this paper, we first propose and develop three different academic networks with a novel data cube based modeling method, and then we perform automated decision processes on these networks. As the first step of the study, we integrate DBLP and CiteSeerX by employing a simple technique called canopy clustering. After the integration of the databases, the modeling stage of the academic networks is performed. In this study, each node as apart from the studies described in the literature is represented by a corresponding data cube with respect to the kind of the network being considered. In order to appropriately analyze the data cube, the OLAP technology is utilized. As the next step of the study, our aim is to automatically find relevant persons, topics and venues from each network. However, it is not an easy task to extract knowledge with low running time and high accuracy from such very huge information networks. In order to overcome this problem, a multi-agent based algorithm is proposed. We evaluate our method with the author network using a benchmark dataset of how well the expertise of the proposed experts matches a given query topic. Our experiments covering other networks show that the proposed strategies are all effective to improve the retrieval accuracy.