Tweedie-Hawkes Processes: Interpreting the Phenomena of Outbreaks

Tianbo Li, Yiping Ke

Abstract

Self-exciting event sequences, in which the occurrence of an event increases the probability of triggering subsequent ones, are common in many disciplines. In this paper, we propose a Bayesian model called Tweedie-Hawkes Processes (THP), which is able to model the outbreaks of events and find out the dominant factors behind. THP leverages on the Tweedie distribution in capturing various excitation effects. A variational EM algorithm is developed for model inference. Some theoretical properties of THP, including the sub-criticality, convergence of the learning algorithm and kernel selection method are discussed. Applications to Epidemiology and information diffusion analysis demonstrate the versatility of our model in various disciplines. Evaluations on real-world datasets show that THP outperforms the rival state-of-the-art baselines in the task of forecasting future events.
BibTeX
@inproceedings{li2020tweedie,
  author = {Tianbo Li and Yiping Ke},
  title = {{Tweedie-Hawkes Processes: Interpreting the Phenomena of Outbreaks}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2020},
  doi = {10.1609/aaai.v34i04.5902},
}
Figure