Extensions and Implementation of Recent Improvements to GLMs

Thorough decision-making requires extensive analysis of current and historical experience. This is done through accurate modeling techniques. In actuarial science and insurance, models are used to evaluate risk and make predictions about future claims using a set of variables also known as predictors. However traditional modeling techniques have shortcomings that can be evaded if we add certain features to the models to improve their predictive ability such as hierarchical interactions. Predictors do not have a direct effect on the dependent variable they come in an interactive manner as well. To account for these, adding hierarchical interactions in the models would improve their predictability. The goal of this project is to update existing R package to create an algorithm that includes hierarchical interactions. This will improve Aviva’s predictive analytics and thus enable the company to make better ratemaking decisions.

Faculty Supervisor:

Jose Garrido

Student:

Partner:

Aviva Canada Inc.

Discipline:

Mathematics

Sector:

Finance and Insurance

University:

Concordia University

Program: