Micro-level insurance claim count modelling: a Cox

SÉMINAIRE DE
MATHÉMATIQUES ACTUARIELLES ET FINANCIÈRES
organisé par Quantact, le Laboratoire de mathématiques actuarielles et financières du CRM
LB-921.04
1400 de Maisonneuve Ouest, Montréal
Pavillon J.W. McConnell (Library) Building, SGW Campus, Concordia University
17 avril 2015, 14:30-15:30
Xinda Yang
School of Risk and Actuarial Studies,
University of New South Wales Business School
UNSW Sydney NSW 2052, Australia
Micro-level insurance claim count modelling: a Cox process approach
Micro-level claims modelling is attracting increasing interest in both academia and the industry. In this paper, we
present a Cox process approach to model insurance claims counts based on micro-level observations, and
implement the model in the prediction of the numbers of Incurred-But-Not-Reported claims. By using a Cox
process approach, our model allows for over-dispersion and serial dependency of claims counts. In order to
allow for practicalities of insurance risks, we further incorporate risk exposure information and allow for delays in
reporting. The performance and application of our model and estimation algorithms are illustrated using both
simulated and read datasets.
Site web: www.quantact.uqam.ca/pages/seminaires
SEMINAR OF
ACTUARIAL AND FINANCIAL MATHEMATICS
organized by Quantact, the CRM Laboratory of Actuarial and Financial Mathematics
LB-921.04
1400 de Maisonneuve Blvd. W., Montréal
Pavillon J.W. McConnell (Library) Building, SGW Campus, Concordia University
April 17 2015, 14:30-15:30
Xinda Yang
School of Risk and Actuarial Studies,
University of New South Wales Business School
UNSW Sydney NSW 2052, Australia
Micro-level insurance claim count modelling: a Cox process approach
Micro-level claims modelling is attracting increasing interest in both academia and the industry. In this paper, we
present a Cox process approach to model insurance claims counts based on micro-level observations, and
implement the model in the prediction of the numbers of Incurred-But-Not-Reported claims. By using a Cox
process approach, our model allows for over-dispersion and serial dependency of claims counts. In order to
allow for practicalities of insurance risks, we further incorporate risk exposure information and allow for delays in
reporting. The performance and application of our model and estimation algorithms are illustrated using both
simulated and read datasets.
Website: www.quantact.uqam.ca/pages/seminaires