Abstract

Predicting frequencies and classifying severities of risks has been a major concern in InsurTech, FinTech, and the emerging field of cyber risk security. When dealing with insurance and finance datasets, some prevalent classifiers cannot comprehensively handle the categorical features present, resulting in a substantial loss of information; others that well handle categorical and discrete features may lack the mechanism to cope with continuous features. Moreover, they do not explicitly account for the relatively strong dependence structures among these feature variables, especially categorical ones.

In this talk, we propose to effectively model implicit strong enough dependence by comonotonicity, and perform risk classification through our newly proposed Comonotone-Independence Bayes Classifier (CIBer), achieving a far better clustering for predictive features. CIBer can thus facilitate an effective classification, and it can be enhanced to serve as a powerful regressor against all types of feature variables. We shall demonstrate the effectiveness and profitability of CIBer among prevalent data analytics tools through empirical studies upon representative datasets in finance and insurance. Finally, we sketch out how its immediate use in cyber risk detection notably improves the prediction power of both claim frequency and severity.

Biography

Prof. Phillip Yam holds a BSc in Actuarial Science with First Class Honours from HKU, an MASt with Distinction from University of Cambridge, and a DPhil from University of Oxford. He was awarded the E. M. Burnett Prize in Mathematics from Cambridge, and junior research fellowship from Erwin Schrödinger International Institute for Mathematics and Physics at University of Vienna. He is now Professor and Director of QFRM programme at Department of Statistics and Data Science in CUHK, and Assistant Dean (Education) at CUHK Faculty of Science. He has been a research fellow at Hausdorff Research Institute for Mathematics in University of Bonn, a Visiting Professor at both Columbia University and UT Dallas, and a Distinguished Visiting Scholar at UNSW Sydney. He is an Elected Member of the International Statistical Institute, and Editor of a top actuarial journal, Insurance: Mathematics and Economics. He has published over 110 articles in actuarial science, applied math, control theory, data analytics, and finance, as well as pioneering monographs on mean field theory and financial data analytics. He has supervised 30+ students and postdocs, many now prominent in academia and industry worldwide.