Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: Ph.D. Dissertation
- Language: Farsi
- Document No: 49614 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Afshin Hemmatyar, Ali Mohammad; Rabiee, Hamid Reza
- Abstract:
- The emergence of online social networks (OSN) has encouraged scientists to investigate them more deeply due to availability of excessive data and increased computational power. Diffusion is a fundamental process taking place in such networks. The spread of news, ideas, influence and diseases are examples of diffusion in social networks. The importance of information diffusion in different disciplines such as economics, politics and social behavior has motivated us to study diffusion networks in this thesis. Diffusion’s behavior is strongly influenced by the underlying network in social networks. Despite this fact most research done in this area ignores the dynamics of the underlying network. In addition, it is mostly presumed to have full knowledge about the underlying network. Although these relaxing assumptions simplify the analysis, the results are useless in real scenarios where the underlying network is dynamic and latent. In this thesis, we investigate the diffusion network extraction problem when the underlying network is both dynamic and latent. We model the diffusion behavior (existence probability) of each edge as a stochastic process and utilize the Hidden Markov Model to discover the most probable diffusion links according to the current observation of the diffusion process, which is the infection time of nodes and the past diffusion behavior of links. We evaluate the performance of our Dynamic Diffusion Network Extraction (DDNE) method, on both synthetic and real datasets. Experimental results show that the performance of the proposed method is independent of the cascade transmission model and outperforms the state of art method in terms of F-measure. A fundamental fact that can provide insights for analyzing and extracting diffusion networks is interdependency between the diffusion behaviors of nodes (links). Interdependency may come from different causes such as the community structure property. We are interested in capturing these interdependencies while extracting the diffusion network. In this regard, we define a more accurate observation metric that considers dependency among nodes in diffusions. Our experiments on real data show that the proposed method improves the extracted diffusion network according to community based metrics. Observing the infection time of nodes in different cascades may not be feasible in cases where we have sampled data, noise or etc. We improve our diffusion network extraction model to overcome the challenge of noisy observation. The idea is to model dependency among nodes’ diffusion behavior as a process over time. The dependency between to nodes is related to their past dependency besides the current observation. Considering the past behavior of nodes (dependency) helps us recover the effect of noisy data while extracting diffusion links
- Keywords:
- Noise ; Hidden Markov Model ; Interdependency ; Diffusion Network ; Information Diffusion ; Social Networks ; Online Dynamic Social Networks ; Partial Observation
Digital Object List
-
محتواي کتاب
- view
Bookmark