Dynamical Processes in Complex Network based on Machine Learning

Abstract

Propagation processes are ubiquitous in many complex network-based systems. The spread of epidemics, labels or information share similar characteristics and depend profoundly on the organization of the network. The complex networks have heterogeneous nature, where some vertices are more influential than others and there are different types of vertices connected to each other. In this way, understanding how the network structure impacts the dynamics and also how to infer the structure from these dynamics is of paramount importance to the area. In this project, we aim to develop methods to improve the tasks of the dynamical processes of machine learning in complex networks, analyzing the impact that the network exerts on them. We will analyze which vertices are most influential in the label propagation task and which can be recommended to be labeled in order to maximize the accuracy results. Also, we will develop methods to detect the community structure of the network from the propagation dynamics. Finally, through the use of multilayer networks, we will develop a method of selection of attributes proper to networks. The analyses will be conducted considering the theory of complex networks and machine learning, using artificial and real databases, evaluating the methods of the literature and applying to possible real problems.