Outlier-Resistant Models for Doubly Stochastic Point Processes

Leo Elsaesser
University of Wisconsin-Milwaukee
MS Graduate Student 

“This thesis proposes an outlier-resistant multiplicative component model for doubly stochastic point processes. The model is based on a principal component decomposition of the log-intensity functions, using heavy-tailed t-distributions for the component scores. As an example of application, the temporal distribution of bike check-out times in the Divvy bike sharing system of Chicago is analyzed using the new model.”

Committee Members:
Profs. Daniel Gervini (Advisor); David Spade, & Chao Zhu