Abstract
dc:descriptionThe emergence of Online Social Networks (OSNs) in the Internet era has provided a new dimension in the opinion diffusion. Online users can share freely and instantly their beliefs with their friends and comment or respond to posts of other users. New terms have arisen stemming from the settings of the modern online landscape. Influencers yield significant impact on other members of their community and it comes as no surprise that major companies have developed marketing strategies that utilize the properties of the most influential nodes in OSNs in order to promote their products and services. Recently, disturbing phenomena of misinformation and manipulative bias have occurred that kickstarted a vivid debate about the role of OSNs in modern societies.The structural characteristics of OSNs and the activity of their online users offer solid ground for research and experimentation. A significant number of scientific efforts focus on the analysis of the influence yielded to online users based on their relations with other members of the community and their activity. The theories and models proposed by the field of opinion dynamics serve as a valuable tool that supports the researchers’ endeavors. Many algorithmic approaches of the influence analysis base their incentives on the findings and formulations of opinion diffusion models in order to tackle their challenging task. In this thesis, we utilize the mathematical formulation of a prominent opinion diffusion model, the DeGroot model, in order to examine the mechanism of influence in social graphs. Our motivation stems from the need to investigate the range of intervention that can be achieved by strategic entities and the impact of their manipulative approach on the community. The mathematical formulation of the DeGroot model enables us to define influence scenarios and examine their behavior on a wide range of social graphs through model checking techniques. In order to produce realistic scenarios that can be applied on real – life cases, we develop intervention schemes that incorporate the activity of a single or multiple users in the social network. A variation of these scenarios entails the limitations of complex settings where the strategic entities are constrained in the construction of an optimal strategy. Experimental results confirm that our approach is applicable on large social networks although, in the case of the constrained settings, the analysis exhibits poor performance due to the increased complexity of the scenario.Furthermore, we develop a methodology based on the DeGroot model with stubborn agents for the analysis of complex real – life graphs generated by modern OSNs in order to extract valuable information from their structural characteristics. The application of our approach on the Greek news media and political scene confirms our hypothesis that the structure of OSNs suffice for the estimation of vague characteristics such as the political affinity and orientation of online nodes of interest. The efficacy of the proposed technique is undoubtably supported by our endeavor to approximate an elaborate feature such as the political impartiality of news media. The case study of the Greek news media and political scene of April 2018 provides clear evidence that our algorithmic approach can answer bewildering questions such as: “Which is the most impartial news media in a political scene”.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Gyftopoulos, Sotirios
- Γυφτόπουλος, Σωτήριος
Subjects
dc:subject × 18- Κοινωνικά δίκτυα
- Επιρροή
- Μέσα κοινωνικής δικτύωσης
- Μοντέλο DeGroot
- Στοχαστικές διεργασίες
- Ανάλυση επιρροής
- Social networks
- Influence
- Online social networks
- DeGroot model
- Stochastic processes
- Influence analysis
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Υπολογιστές, Υλικό (hardware) και Αρχιτεκτονική
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Computer science, Hardware and Architecture
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/49594