Financial/Actuarial Mathematics Seminar

Fall 2004: Thursdays 3:10-4:00, 3088 East Hall



Compound Poisson disorder problem

Savas Dayanik

Princeton University, Operations Research and Financial Engineering

December 9, 2004



Abstract

Suppose that the mean measure of a compound Poisson process changes from one measure to another at an unknown and unobservable time. The problem is to detect this change-time as quickly as possible based on the historical observations of the point process. In this talk, we shall describe an explicit solution of this problem and give some numerical examples. In insurance and manufacturing, quickest change-detection problems arise from the need to adapt optimal decisions to changing regimes when the changes are not directly observable. This is a joint work with Semih O. Sezer from Princeton University.


Back to the schedule