Functional data analysis (FDA), which deals with data that are collected as curves rather than multivariate data points, has received increasing interest in statistics community as well as various areas of applied statistics, such as psychology, medicine, machine learning and engineering. Its applications in Actuarial Science, however, are still at early stage. Existing applications include functional time series analysis for forecasting mortality and fertility rates, functional principal component analysis for estimation of risk of claim occurrence in vehicle insurance and in analysis of cohort life tables.
The aim of the proposed project is to further investigate the applications of FDA in Actuarial Science and to develop statistical methodologies and theories suitable for the problems arising from this area, focusing particularly on mortality and risk selection. For example, the project will contribute to study of dynamic relationship between mortality rates and other factors, discovery of the most important factors affecting mortality, heterogeneity in life tables, prediction of long term trend of mortality, modelling and prediction of claim occurence, clustering of different relationships between claim risk and other factors (nutrition, education, social class), etc.