Identification of Investment Decision Process Using Online Inverse Optimization
We consider asset allocation as an optimal decision process governed by Markowitz Framework and Black-Litterman model. Under this framework, several important decision variables, such as risk aversion, expected portfolio benchmark return, and expected asset return, can all be learned by inverse optimization process. In this talk, we investigated a novel online inverse optimization framework on portfolio data, market news and asset price data to identify decision variables in investment decisions. We showed how those learned variables can be used to understand investor’s decisions, fund managers’ advices and recommendations, and ultimately enhance automated financial advice capabilities.
Shi Yu is a scientist who is enthusiastic about applying machine learning techniques to solve challenging problems. His main interests are Natural Language Processing, Large-scale Machine Learning, GPU/MPI Parallel Computing, and interesting problems in Finance and Insurance Industry. He is a Principal data scientist at Vanguard Group overseeing data science initiatives for Vanguard IIG, FAS and International. Shi holds a Master’s degree in Artificial Intelligence and a PhD in Electrical Engineering from K. U. Leuven. He finished his post-doc research at University of Chicago, and worked at IBM, Deloitte and American Family Insurance before joining Vanguard.