2019年11月8日下午，应广州大学经济与统计学院和岭南统计科学研究中心的邀请，在行政东前座412会议室，香港城市大学严兴博士作了题为“Machine Learning for Quantitative Tail Risk Management”的学术报告——暨“羊城讲坛”第六十二讲，旨在进一步提高年轻学者及研究生对相关研究的理解。此次讲座由胡建明老师主持，相关专业的师生参加了此次讲座。
报告的主要内容：Extreme events happen more frequently than we think. Forecasting the probability of extreme events in financial markets and the following decision making, which we call tail risk management, is extremely important for investments and regulations. In our research, we design novel machine learning models that capture not only the tail risk of an individual asset but also tail dependence among multiple assets. For individual asset return, we design a parametric quantile function for heavy tail modeling and combine it with an LSTM neural network to model the conditional distribution and forecast quantiles. For multiple asset returns, we propose a transformed random vector to separately model their tail dependencies from the correlations. We do adequate numerical experiments to verify our models’ ability for forecasting tail risk. Besides, we have interesting findings on tail behaviors and they should have important implications for asset pricing.