Academic

Influential entities can occur in social media, financial markets, energy transmission, you name it. In the presence of short time series while the system is high-dimensional, an estimation method not accounting for influential entities will be inefficient. In this talk we introduce methodologies to improve said estimation for the returns and extreme values of asset networks. We further extend the methodologies to risk in energy transmission systems and influencer detection in social media. We show that in the presence of influential entities, the methodologies provide higher accuracy, better forecasting results, and improve the understanding of the underlying systems.