Squirrel search algorithm for portfolio optimization

Abstract

Portfolio Optimization is a standard financial engineering problem. It aims for finding the best allocation of resources for a set of assets. This problem has been studied and different models have been proposed since the classical Mean-Variance model was introduced by Harry Markowitz in 1952 and the later modified version by William Sharpe. The inclusion of real-life constraints to the problem has led to the introduction of the extended Mean-Variance model. However, the successes of nature-inspired algorithms in hard computational optimization problems have encouraged researchers to design and apply these algorithms for a variety of optimization problems. In this paper, we design and adapt a Squirrel Search Algorithm (SSA) for the unconstrained and constrained portfolio optimization problems. SSA is a very recent swarm intelligence algorithm inspired by the dynamic foraging behavior of flying squirrels. The proposed SSA metaheuristic approach is compared with a variety of approaches presented in the literature such as classical single metaheuristics, hybrid metaheuristic approaches and multi-objective optimization approaches for portfolio optimization. Comparative analysis and computational results using different performance indicators show the superiority of the proposed approach for the unconstrained portfolio optimization using both extended Mean-Variance and Sharpe models. For the constrained version of the problem, the proposed approach has also achieved highly competitive results for the different models adopted.