To answer questions that look beyond individual entities, network science studies the interactions taking place in complex systems. These complex systems can often have temporal information on when an interaction occurs. Incorporating such information in an analysis can lead to richer insights and allow us to study the systems over time. However, the incorporation of temporal information is not straightforward and requires careful consideration of the methods applied.
In this thesis we study two complex systems over time and introduce new temporal algorithms for efficient processing. First, we study the Dark Web cryptomarket Evolution, a combined forum and marketplace where drugs, and ID/credit information were illegally traded. We study the communication network on the forum to predict which users are most likely to be (successful) vendors on the market, both now and in the future. We find that topic engagement, a measure of response activity to...
Show moreTo answer questions that look beyond individual entities, network science studies the interactions taking place in complex systems. These complex systems can often have temporal information on when an interaction occurs. Incorporating such information in an analysis can lead to richer insights and allow us to study the systems over time. However, the incorporation of temporal information is not straightforward and requires careful consideration of the methods applied.
In this thesis we study two complex systems over time and introduce new temporal algorithms for efficient processing. First, we study the Dark Web cryptomarket Evolution, a combined forum and marketplace where drugs, and ID/credit information were illegally traded. We study the communication network on the forum to predict which users are most likely to be (successful) vendors on the market, both now and in the future. We find that topic engagement, a measure of response activity to topics by users, is the best predictor of vendor success. Furthermore, we find that betweenness centrality, a measure of how well a user connects various parts of the network, is able to find successful vendors with low activity, thereby compensating for a weakness of topic engagement. Additionally, high betweenness centrality can be an early warning signal of future vendor success.
Second, we study the scientific publication system. We study both the evolution of the city collaboration network over time, finding it to have become more stable over time, as well as the effects of persistence and freshness on the citation success of teams. Persistent teams are identified based on temporal cliques, for which a new efficient algorithm is introduced. These persistent teams are shown to be overrepresented on highly cited publications and to be more likely to produce such highly cited publications early in the collaboration. Freshness impulses, from new collaborations started by team members, are shown to sustain success, while persistence impulses, from preceding collaborative experience by team members, are shown to facilitate early success. Thus, freshness and persistence each contribute to the success of a team, but at different stages of their lifespan.
Show less
