Dinâmica de temas abordados no Twitter via evolução de clusters

Abstract

The massive use of social networks in recent years has produced a large volume of information generated by its users, who often share their feelings, opinions about major events, disasters, epidemics, product launches, among other events. Understanding how specific subjects are evolving in social networks and how they are related each other, it can be relevant for organizations that aim to make better marketing strategies, advertising targeting, get feedbacks events or any released product. However, to analyze this large volume of nonautomated data consists of a non-trivial problem. In the context of this work, we apply a clustering technique to identify sets of issues in Twitter, and we also apply cluster evolution to discover how these issues are dynamic and the transition to each other. We aim to understand and discuss about such cluster evolution. In addition, apply two similarity measures the clustering of data collected and analyzed similarities and differences in the results obtained by both. In this work, we detected the following transitions for clusters: Creation, Survival, Disappearance, Merge, Split, Expansion and Shrinkage of clusters.