Robust Metaheuristic Optimization for Algorithmic Trading: A Comparative Study of Optimization Techniques

Algorithmic trading heavily relies on the optimization of rule-based strategies to maximize profitability and ensure robustness under volatile market conditions. Traditional optimization methods often face limitations when dealing with the nonlinear, high-dimensional, and dynamic nature of financial search spaces. This study introduces a Metaheuristic-based framework for financial strategy optimization that focuses on the modeling and resolution of the problem through population-based search algorithms. The framework evaluates four Metaheuristic optimization techniques within a unified design, enabling a consistent and fair comparison of their performance in optimizing trading rules. To ensure realistic and time-consistent evaluation, the experimental setup incorporates a Rolling Windows Validation approach, allowing the assessment of model performance across successive market periods. Beyond improving convergence behavior, Diversity is employed as a metric to assess the quality and exploration capability of the search process, providing deeper insight into algorithmic performance. Experimental results, obtained from real market data, demonstrate substantial improvements in profitability consistency and risk-adjusted performance compared to conventional optimization approaches. The findings confirm that Metaheuristic optimization offers a robust and flexible alternative for the design and refinement of algorithmic trading systems in complex and dynamic financial environments. Interestingly, Differential Evolution exhibited persistently high Diversity, suggesting the presence of multiple distant yet competitive optima in the financial search space, where functional convergence coexists with geometric dispersion.