Utilization of Artificial Intelligence for Portfolio Optimization in Shallow Markets

Introduction: Portfolio optimization remains a compelling research challenge despite advances in techniques and artificial intelligence. This complexity largely stems from the difficulty in accurately identifying and modeling dynamic market factors. Variables such as interest rates, inflation, economic growth, and foreign exchange rates fluctuate continuously, making portfolio construction and adjustment uncertain. As a result, portfolio strategies often rely on correction mechanisms that may be imprecise or insufficient under volatile and structurally complex market conditions. Objectives: The objective of this study is to compare momentum-based algorithmic trading with buy-and-hold strategies in shallow markets, assessing AI’s ability to exploit inefficiencies and enhance portfolio performance. Methods This study applies artificial intelligence optimization methods, including momentum strategies enhanced by Support Vector Machines (SVM), to construct actively managed portfolios. A sample of 10 equally weighted stocks is selected and rebalanced over a rolling 252-day window using adjusted closing prices retrieved from Yahoo Finance via STATA. The algorithm dynamically adjusts holdings based on market signals. The performance of the optimized algorithmic portfolios is compared against a traditional buy-and-hold strategy. Statistical significance of return differences is assessed using paired t-tests to evaluate the effectiveness of optimization in shallow and less efficient market conditions. Results: The results show that the trading algorithm outperformed the buy-and-hold strategy in 18 out of 31 cases, while the buy-and-hold strategy outperformed in 13 cases. On average, the algorithm achieved higher returns in 51.85% of the observations. However, the difference in mean returns between the two strategies is not statistically significant at the 5% level, according to the paired sample t-test. This suggests that while the algorithm showed slightly better performance, the results do not provide strong statistical evidence of its superiority. Conclusions: Portfolio optimization in shallow markets presents unique challenges due to limited liquidity and weaker market efficiency. While AI can exploit mispricing and enhance returns through active strategies, data flaws and structural constraints complicate its effectiveness. Competing algorithms, timing issues, and behavioral factors further intensify complexity in these less developed markets.