We present a novel deep learning methodology to gain pragmatic insights into the behavior of an insured person using unsupervised variable importance. It lays the groundwork for understanding how insights can be gained into the fraudulent behavior of an insured person with minimum effort. We propose a new variable importance methodology incorporated with two prominent unsupervised deep learning models, namely, the autoencoder and the variational autoencoder. Each model's dynamics is discussed on how models can be adapted for fraud detection and how results can be perceived appropriately. Both qualitative and quantitative performance evaluations are conducted. To broaden the scope of reference of fraud detection setting, various metrics are used in the qualitative evaluation.

Hailiang Yang, ASA and HonFIA, received his PhD degree from the University of Alberta and his Master degree in actuarial science from the University of Waterloo. He is currently a professor in the Department of Financial and Actuarial Mathematics at Xian Jiaotong-Liverpool University (XJTLU). Before joining XJTLU, he worked at the University of Hong Kong for 27 years. His research focuses on actuarial science and mathematical finance. He has worked with many leading figures in the field. He has supervised more than 20 PhD students, and his graduate students are, in many cases, now well-known researchers in their own right. Furthermore, he is an editor of Insurance; Mathematics and Economics and an associate editor of six other journals. In addition, he is an Associate of the Society of Actuaries, and he was elected as an Honorary Fellow of the Institute and Faculty of Actuaries and a Corresponding Member of the Swiss Association of Actuaries in 2014. He is an elected member of the International Statistical Institute (ISI). He received an Outstanding Research Award from the University of Hong Kong in 2013–2014.