Seminar in applied mathematics and statistics
Online via the following Zoom link:
https://ucph-ku.zoom.us/j/65811603956?pwd=b2t2MGs5NGduRVRDcVVMVWFDTE5wZz09
SPEAKER: Mikko Pakkanen (Department of Mathematics, Imperial College)
TITLE: State-dependent Hawkes processes: Theory, Estimation and Applications
ABSTRACT: Hawkes processes are a class of point processes that can describe event-by-event data that are clustered in time, in the sense that each event increases the likelihood of another event occurring shortly afterwards; a property called self-excitation. Hawkes processes have been used to model earthquakes and aftershocks, social media activity, neuronal firing, epidemics and financial data, for example. In my talk I will introduce a new, state-dependent version of the Hawkes process, where the point process and its self-excitation behaviour is modulated by a state process which, reciprocally, is influenced by the past evolution of the point process. Such a process can capture any variation in the levels of self-excitation that is explainable by observable, endogeneous "features". I will present basic theory on the existence of such processes, their maximum-likelihood estimation and finally an empirical application to high-frequency financial data.
Joint work with Maxime Morariu-Patrichi.