Lightweight Optimization of Investment Portfolios: An Empirical Study Based on the SPEA2 Multi-Objective Evolutionary Algorithm

Abstract

Portfolio optimization is a classical yet computationally complex problem in computational finance, particularly when multiple conflicting objectives, such as maximizing return and controlling investment risk, are considered simultaneously. Many existing evolutionary and interactive optimization approaches can generate high-quality portfolio solutions; however, their long runtime and high computational cost limit their practical use in resource-constrained or time-sensitive decision-making environments. This paper presents a lightweight portfolio optimization framework that reduces computational complexity by deliberately limiting the search space through the selection of a small but representative set of assets. The proposed framework employs the Strength Pareto Evolutionary Algorithm 2 (SPEA2) as the main optimization engine, owing to its Pareto-based selection mechanism and external archive structure, which support effective convergence while maintaining lower computational overhead. Experiments were conducted on a dataset consisting of 14 selected stocks over a monthly investment horizon. The performance of SPEA2 was compared with the widely used Non-dominated Sorting Genetic Algorithm II (NSGA-II) under identical experimental settings. The results show that SPEA2 achieved a final portfolio value of $1,847,560.42, compared with $1,775,711.48 for NSGA-II. In addition, SPEA2 completed the optimization process in 8,041 seconds, whereas NSGA-II required 160,809.30 seconds, making SPEA2 approximately 20 times faster. The combined time-return efficiency also favored SPEA2, with $105.46 profit per second compared with $4.82 for NSGA-II. These findings indicate that SPEA2, when combined with controlled dimensionality reduction, can provide a practical and efficient approach for portfolio optimization under computational and time constraints.