Portfolio Optimization of Sharia-Based Stocks Using Deep Learning: A Case of Bahrain Bourse

Optimizing portfolios consists of finding the most efficient allocation of funds amongst selected assets. Since the advent of technology in recent years, the process of optimizing such portfolios has undergone substantial evolution driven by integrating machine learning and incorporating complex problem elements to more accurately reflect real-life scenarios. This paper studied solving such a problem using genetic algorithm based on downside mean-variance model that was applied on Sharia-complaint stock listed on Bahrain Bourse. Researchers used the daily closed stock prices between April 2015 and April 2023 to calculate the daily returns and fed them into the algorithm to find the optimal fund allocations amongst the 13 stocks of the Bahrain Islamic Index (BIX). The resulted optimized portfolio performance was then compared with the BIX and mean-variance portfolios using the cumulative daily returns along with Sharpe and Sortino ratios, which all confirmed that the study's model offered higher daily returns for one unit of risk. The results also revealed that the performance of the study's portfolio outperformed the others during COVID-19 pandemic which proved its ability to mitigate systematic risk caused by global crises.