Black-Litterman Portfolio Optimization using Machine-Learning, Deep Learning and Reinforcement Learning Algorithms

Abstract

Shigolakov Ivan Vasilevich and Joe Wayne Byers

A portfolio optimization plays a critical role in the financial world. Today the portfolio managers, financial and investment analysts, quantitative researchers from different financial institutions and companies try to find new or better approaches to optimize the investment portfolio in comparison with classical techniques. Nobel prized Harry Markowitz in 1952 introduced his world famous model which is still used by financial scientists in researches as a benchmark or as a background to develop new approaches. The model has some limitations in terms of data normality assumption, historical mean- variance approach instead of potential risks evaluation and others. As a result, some assumptions become irrelevant in real market conditions. Black-Litterman model developed by Fischer Black and Robert Litterman in 1992 attempts to handle some limitations of Markowitz model. This research project will use the Black-Litterman model as a background for further investigation. The view matrix construction will use the Machine Learning and Deep Learning models predictions. MLDL-based Black-Litterman model is expected to demonstrate higher cumulative returns and Sharpe ratios in comparison with classical Markowitz and BL models.

PDF