Portfolio Optimization Based on Machine Learning

Abstract

Machine learning is utilized widely in portfolio prediction and optimization, which provides a more reliable results with mean variance optimization. Besides, socially responsible investment portfolio considering social, environment and governance becomes an emerging theme in academic and industrial circles. Basing on both of that, this study aims to use machine learning models to predict return of ESG stocks in American stocks market and mean variance optimization for creating portfolio. The consequence demonstrates SVR has less residual in prediction and maximum sharpe or minimum variance portfolio shows similarly. Both portfolios get closing return and better resilience during market undertaking pressure comparing with S&P 500. This research confirms that it is useful to predict by machine learning and ESG assets may bring extra return in the bearish market.