Portfolio Selection and Stock Returns: The Role of Machine Learning Algorithms in Asset Choices

Abstract

The investigation has been made across various financial assets available globally to develop a portfolio that may generate higher returns with low risk and minimal drawdowns and even may perform well under stressful financial happening. The main objective of this study is to provide a strategically effective weighting process to build an efficient and optimally diversified portfolio by using various categories of financial assets classified into three samples. The machine learning algorithms have been applied to diversify and build a suitable portfolio by evaluating the suitability of the rebalancing approach. The last twenty years' daily data of various assets permits analysis of the dynamic behaviors and underlying patterns of the assets as a part of the portfolio. Although the long-run variations in asset returns are not an effective way to make good forecasts about their behavior in the future this makes it possible to test their resilience and response during financial stress-events. Findings suggest that portfolio diversification should incorporate some top-position assets from the main categories of financial markets to classify the ranking of various assets and assigning equal weights to each asset in the portfolio ensures remarkably high returns.