Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 47566 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Rabiei, Hamid Reza
- Abstract:
- Nowadays, many entities and the relationships between them create different types of networks. Graphs are appropriate tools to model these networks. In a graph, nodes show individuals and edges show the relations between them. One of the most important research problems in the field of network analysis is community detection. In a network, a community is a group of nodes that has a lot of connections to the nodes inside the same group and a few to the ones which are outside. Community detection has many real world applications. Recommending items in recommender systems, detecting spy and terrorist groups and predicting future links between members in a social network are some examples where community detection algorithms are incorporated. The community detection problem has been investigated using different approaches. Traditional methods assumed that every node belongs to exactly one community (disjoint groups). But, in real world, a node can belong to many groups, so the concept of overlapping communities is proposed. Furthermore, due to huge size of networks, estimation of the true number of communities is too difficult and the assumption that it is known a priori is far from reality. On the other side, in many community detection studies, the network under investigation is considered static, which means they assumed the network is fixed and does not change. However, the networks change constantly over time and members and connections are added and removed. Considering network evolution over time increases the precision of detected groups and makes them more realistic. In this dissertation, we have studied the overlapping community detection problem in dynamic networks, without any prior knowledge about the number of communities. Dynamic network in our work means a set of network snapshots in discrete time intervals that all changes during a time interval are aggregated at the end of it. In the first step, we have proposed a generative model for a static network based on link communities, afterward we extract the group memberships of nodes by applying statistical inference methods on observed graph. The proposed method is able to discover overlapping communities and the number of them simultaneously, in polynomial time, using chinese restaurant process (CRP), as a prior knowledge and the adjacency matrix as observations. The experimental results on synthetic and real datasets show the superiority of our static method over other community detection methods. In the second step, the proposed method in the first step is extended to dynamic networks to discover consistent communities over time. Our dynamic method is based on a non-parametric dynamic model which is called recurrent chinese restaurant process (RCRP). The proposed method is able to determine the number of communities and the consistent communities in every snapshot of network knowing the history of network and communities. In the experimental results, we compared the proposed method against four other methods on 13 different datasets with different types of community evolution. The results demonstrate that the proposed method can extract the true number of communities, capture different types of the community evolution and preserve accuracy of discovered communities during time snapshots
- Keywords:
- Social Networks ; Bayesian Nonparametric Model ; Dynamic Mesh ; Overlapping Networks ; Recurrent Chinese Restaurant Process
Digital Object List
محتواي کتاب
Bookmark