Vertices in a real-world social network can be grouped into densely connected communities that are sparsely connected to other groups. Moreover, these communities can be partitioned into successively more cohesive communities. Despite an ever-growing pile of research on hierarchical community detection, existing methods suffer from either inefficiency or inappropriate modeling. Yet, some cut-based approaches have shown to be effective in finding communities without hierarchies. In this paper, we study the hierarchical community detection problem in large networks and show that it is NP-hard. We then propose an efficient algorithm based on edge-cuts to identify the hierarchy of communities. Since communities at lower levels of the hierarchy are denser than the higher levels, we leverage a fast network sparsification technique to enhance the running time of the algorithm. We further propose a randomized approximation algorithm for information centrality of networks. We finally evaluate the performance of the proposed algorithms by conducting extensive experiments using real datasets. Our experimental results show that the proposed algorithms are promising and outperform the state-of-the-art algorithms by several orders of magnitude.

Real social networks contain many communities, where members within each community are densely connected with each other, while they are sparsely connected with the members outside of the community. Since each member can join multiple communities simultaneously, communities in social networks are usually overlapping with each other. How to efficiently and effectively identify overlapping communities in a large social network becomes a fundamental problem in the big data era. Most existing studies on community finding focused on non-overlapping communities based on several well-known community fitness metrics. However, recent investigations have shown that these fitness metrics may suffer free rider and separation effects where the overlapping region of two communities always belongs to the denser one, rather to both of them. In this paper, we study the overlapping community detection problem in social networks that not only takes the quality of the found overlapping communities but also incorporate both free rider and separation effects on the found communities into consideration. Specifically, in this paper, we first propose a novel community fitness metric - triangle based fitness metric, for overlapping community detection that can minimize the free rider and separation effects on found overlapping communities, and show that the problem is NP-hard. We then propose an efficient yet scalable algorithm for the problem that can deliver a feasible solution. We finally validate the effectiveness of the proposed fitness metric and evaluate the performance of the proposed algorithm, through conducting extensive experiments on real-world datasets with over 100 million vertices and edges. Experimental results demonstrate that the proposed algorithm is very promising.

Recent studies show that individuals in a social network can be divided into different groups of densely connected communities, and these individuals who bridge different communities, referred to as structural hole spanners, have great potential to acquire resources/information from communities and thus benefit from the access. Structural hole spanners are crucial in many real applications such as community detections, diffusion controls, viral marketing, etc. In spite of their importance, little attention has been paid to them. Particularly, how to accurately characterize the structural hole spanners and how to devise efficient yet scalable algorithms to find them in a large social network are fundamental issues. In this paper, we study the top-k structural hole spanner problem. We first provide a novel model to measure the quality of structural hole spanners through exploiting the structural hole spanner properties. Due to its NP-hardness, we then devise two efficient yet scalable algorithms, by developing innovative filtering techniques that can filter out unlikely solutions as quickly as possible, while the proposed techniques are built up on fast estimations of the upper and lower bounds on the cost of an optimal solution and make use of articulation points in real social networks. We finally conduct extensive experiments to validate the effectiveness of the proposed model, and to evaluate the performance of the proposed algorithms using real world datasets. The experimental results demonstrate that the proposed model can capture the characteristics of structural hole spanners accurately, and the structural hole spanners found by the proposed algorithms are much better than those by existing algorithms in all considered social networks, while the running times of the proposed algorithms are very fast.

Recent studies have shown that in social networks, users who bridge different communities, known as structural hole spanners, have great potentials to acquire available resources from these communities and gain access to multiple sources of information flow. Structural hole spanners are crucial in many applications such as community detections, diffusion controls, and viral marketing. In spite of their importance, not much attention has been paid to them. Particularly, how to characterize the structural hole spanner properties and how to devise efficient yet scalable algorithms to find them are fundamental issues. In this paper, we formulate the problem as the top-k structural hole spanner problem. Specifically, we first provide a generic model to measure the quality of structural hole spanners, by exploring their properties, and show that the problem is NP-hard. We then devise efficient and scalable algorithms, by exploiting the bounded inverse closeness centralities of vertices and making use of articulation points of the network. We finally evaluate the performance of the proposed algorithms through extensive experiments on real and synthetic datasets, and validate the effectiveness of the proposed model. Our experimental results demonstrate that the proposed model can capture the characteristics of structural hole spanners accurately, and the proposed algorithms are very promising.

The importance of online information retrieval systems has dramatically increased through considerable growth in the size of the web, and the challenges beyond this topic have become a center of attention for many researchers. This remarkable growth in the size of the web has led to in-depth studies on every element of information retrieval systems such as web page ranking algorithms. The accuracy of these algorithms plays a critical role in the search engines, whereas the ranker is responsible for accuracy. Thus, the ranker is a principal module of each search engine. In this paper, a new framework based on web graph and content similarity is presented in order to improve the accuracy of PageRank. This framework is implemented using Jaccard index and cosine similarity measures, and as a result of our empirical analysis, we shall show that putting page content similarity in action increases the accuracy of web ranking in some candidate ranking algorithms. In addition, time complexity and implementation issues are discussed to achieve a practical result.