Comparison of Metaheuristic Algorithms in Portfolio Optimization: Evidence on Hedge Fund Returns

Abstract

This study extends prior research on Artificial Bee Colony (ABC) portfolio optimization by conducting a comparative assessment of several metaheuristic techniques applied to the hedge fund portfolio optimization problem. Following a survey of existing research, we implement five distinct nature-inspired algorithms: two rooted in swarm intelligence (Artificial Bee Colony and Particle Swarm Optimization) and three based on evolutionary principles (Genetic Algorithm, Differential Evolution, and Harmony Search). The asset universe consists of ten indices representing diverse hedge fund strategies, with the optimization objective being the minimization of Conditional Value-at-Risk (CVaR). The efficacy of the resulting portfolios was evaluated against standard industry proxy indices. The empirical validation was conducted over a three-year period, incorporating an annual rebalancing protocol. Our analysis reveals that metaheuristic-optimized portfolios can achieve highly competitive risk-return profiles. Notably, the portfolio constructed via the ABC algorithm delivered performance comparable to a diversified fund-of-funds benchmark, while demonstrating a considerable performance advantage over an equal weighted hedge fund index. These findings affirm the practical utility of metaheuristic frameworks for sophisticated asset allocation, offering a robust methodology for constructing portfolios with managed downside risk in the alternative investment domain, characterized by non-normal return distributions.