Loading...

Diffusion in Social Networks Based on Partial Information

Ramezani, Maryam | 2024

0 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 57580 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rabiee, Hamid Reza
  7. Abstract:
  8. During the last decade, the study of various aspects of social networks has received in- creasing attention. Communication and information propagation among network mem- bers are important issues in social networking. The diffusion process is a fundamental mechanism by which information, behaviors, or new ideas spread over the network. This process starts from a small group of nodes and continues until the majority of members are affected. Although it is possible to observe the times when nodes become infected, de- termining who infected each node is often difficult to ascertain. Analyzing and modeling the diffusion process has various applications in marketing, politics, and social studies. Most of the current diffusion models consider only the active nodes in the network and assume that a node can get infected immediately after exposure, without accounting for delayed behaviors. In this dissertation, the primary objective is to address issues related to information diffusion in social networks while considering partially observation. This challenge is approached from two perspectives. From one perspective, missing data is interpreted as a widespread event, where the joint inference of user interactions and diffusion activities with access to incomplete data is first proposed. Then, assuming the lack of access to user interactions, the dissertation aims to infer the underlying network while preserving its structural properties. From another perspective, missing data can be equated with having limited access to the initial portion of the diffusion data, which introduces challenges in classification, modeling, and prediction under such conditions. The early detection of whether news streams are fake or real, the use of multi-modal data sources for fake news detection, the identification of viral social events, and the modeling of various types of news diffusion are some of the issues addressed from the second viewpoint despite the data gaps. In each section, along with presenting innovative models, the effectiveness of the proposed methods is evaluated through theoretical analysis and experiments on various synthetic and real datasets
  9. Keywords:
  10. Social Networks ; Information Diffusion ; Network Inference ; Partial Information ; Fake News Detection ; Virality ; Diffusion Modeling ; Viral Social Events

 Digital Object List

 Bookmark

...see more