Spatiotemporal Data Analytics based on Complex Networks

Abstract

Complex network theory has helped to identify valuable information in many domains, where systems are complex with non-trivial connections and properties. 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 analyze the patterns of activity and the evolution of dynamic networks. The project includes the design, development, and testing of new techniques for mining temporal and dynamic networks and, the application of network analysis techniques to real-world problems. We will analyze and assess dynamic network patterns in climate data, which are geo and temporally tagged. We are proposing a set of spatiotemporal data analytic methods that, although tested on this climate data, could in principle extended to any temporal data set. Possible other data sets that we will explore are the diffusion of disinformation, voting network data, and dynamic opinion networks. As for grounding for the geo-temporal assessment, we aim to begin by applying and adapting traditional network measures including centrality measures, community detection algorithms, and graph metrics for assessing structure. We will conduct the analyses considering the theory of network sciences, machine learning, and dynamic networks, using artificial and real data sets, evaluating the methods of the literature and applying to real-world problems.