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: 42600 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Rabiei, Hamid Reza
- Abstract:
- The diffusion process propagates information, viruses, ideas, innovations and new be-haviours over social networks. Adopting a new behaviour, which is mentioned as an in-fection, starts from a little group of people. Spreading it over more neighbors and friendscan result in an epidemic phenomenon over the network. As this infection propagates, aninformation cascade will be generated. The spread of information cascades over social net-works forms the diffusion networks. Although observing the infection time of a person ispossible, determining the source of infection is usually a difficult problem. Additionally, inmany applications we can not observe the underlying network which diffusion occurs overit. Therefore, extracting diffusion network has attracted great attention as an emerging areaof multi-disciplinary research.In this work, we propose a scalable algorithmcalled “DNE” (Diffusion Network Extraction)that utilizes the time-series data to extract the diffusion links without assuming any priorknowledge of the underlying network structure and its topological features. We first modelthe spreading information cascades over the network as a Markov randomwalk process. Con-sidering some concepts of randomwalk leads us to a newparameter we call the reaching time.Using this parameter, we define a probabilistic model for extracting diffusion network thatleads to a non-linear computation time. To alleviate this problem, we introduce a tractableapproximation model for this parameter that is independent from the cascade transmissionmodel. This independence can be useful in real networks where the actual model of infor-mation transmission is not available. Utilizing both synthesis and real datasets, we show thatour method can run in linear time with respect to the number of the edges in the network The proposed model can speed up the extraction process up to 300 times with respect to thestate of the art method
- Keywords:
- Information Cascade ; Social Networks ; Information Diffusion ; Diffusion Network ; Markov Random Walk (MRW) ; Underlying Network
Digital Object List-
محتواي پايان نامه - view
Bookmark