Characterizing multiple interactions in dynamic attributed networks based on social concepts

• Main Author: Thiago Henrique Pereira Silva
• Publication Date: 2020
• Format: Doctoral thesis
• Language: eng
• Source: Repositório Institucional da UFMG
• Download full: http://hdl.handle.net/1843/38080
• topic: Social Networks; Dynamic Attributed Networks; Social Computing; Computação – Teses.; Reds sociais – Teses.; Redes de relações sociais – Teses.; Redes complexas – Teses.; Computação social – Teses.

Characterizing dynamic interactions is currently an important issue when analyzing complex social networks. Based on the structural autonomy that informs when people are tightly connected to one another with extensive bridge ties beyond them, we reinforce the importance of the network theory paradigm as fundamental for understanding the complexity that involves actors and their relationships. In this regard, we discuss how to model multiple interactions in dynamic attributed networks and propose a classification method that classifies nodes and dynamic edges based on node-attribute relationships. As a result, it captures the strength of social interactions and how knowledge is transferred across the network. Then, we unveil and illustrate the differences of social interactions in different academic social networks and Q&A communities. Based on the strategic positioning of a particular actor in a social structure, we statistically validate our proposed strategy by means of network properties. Moreover, we perform a sensitivity analysis by stressing it in terms of its robustness to deal with aspects of time, discriminative power of attributes and random scenarios. Finally, we propose unsupervised and supervised strategies that apply our method to identify influential nodes in a social structure, which outperform traditional network metrics and other social-based algorithms.