2021/05/19 by Haoran Wang, Wang, Haoran, Shi Ming Yu +1 · 1 citation
Decision Sciences · Economics, Econometrics and Finance · #91G10 #Advanced Bandit Algorithms Research #Computational Finance (q-fin.CP) #FOS: Computer and information sciences #FOS: Economics and business #Financial Markets and Investment Strategies #Machine Learning (cs.LG) #Portfolio Management (q-fin.PM) #Stock Market Forecasting Methods
paper · pdf · doi:10.48550/arxiv.2105.09264
openalex publication_date 2021/05/19 · openalex created_date 2021/06/22 · openalex updated_date 2026/07/28
Machine Learning (ML) has been embraced as a powerful tool by the financial\nindustry, with notable applications spreading in various domains including\ninvestment management. In this work, we propose a full-cycle data-driven\ninvestment robo-advising framework, consisting of two ML agents. The first\nagent, an inverse portfolio optimization agent, infers an investor's risk\npreference and expected return directly from historical allocation data using\nonline inverse optimization. The second agent, a deep reinforcement learning\n(RL) agent, aggregates the inferred sequence of expected returns to formulate a\nnew multi-period mean-variance portfolio optimization problem that can be\nsolved using deep RL approaches. The proposed investment pipeline is applied on\nreal market data from April 1, 2016 to February 1, 2021 and has shown to\nconsistently outperform the S&P 500 benchmark portfolio that represents the\naggregate market optimal allocation. The outperformance may be attributed to\nthe the multi-period planning (versus single-period planning) and the\ndata-driven RL approach (versus classical estimation approach).\n