Machine Learning-Driven Portfolio Design: A Hybrid MPT–GA Approach with Advanced Risk Visualization – A Review

Vaibhav K. Herwade, Mahesh S. Salunkhe, S. Patil

Abstract

Portfolios have progressed from the basic principles of modern portfolio theory (MPT) towards cutting-edge machine learning (ML)-based models but still suffer considerable deficiencies concerning computational intelligence incorporation into finance. This review article analyzes the progress in creating and validating an MPT-GA-based hybrid model supplemented by advanced risk visualization. In other words, this review attempts to analyze how computational intelligence can be incorporated into theoretical finance. More than forty scholarly works were considered, with the emphasis on the limitations of classical MPT: assumption about the normal distribution, constant correlation coefficients, disregard for transaction costs and behavioral factors, among others. In addition, this work discusses how machine learning can improve the performance of portfolio design by predicting returns and estimating risks. Also, the role of GAs in multi-objective and constrained portfolio optimization is analyzed. At the same time, special attention was paid to the design and validation of advanced visualizations, which will increase the interpretability and effectiveness of portfolio management decisions.

Source: semanticscholar · PDF

Read the AI summary and key takeaways for traders on WOBR Quant Research.