Portfolio optimization is a fundamental problem in financial decision making. One important variation of this problem is index tracking, where the goal is to construct a portfolio that replicates the behaviour of a benchmark index using a limited number of assets. The index tracking problem is computationally challenging due to the combinatorial nature of asset selection and the presence of multiple constraints.

In this paper we propose a genetic algorithm for solving the constrained index tracking portfolio optimization problem. The proposed approach incorporates a dual-population structure combined with elitism, roulette-wheel selection, two-point crossover and adaptive Gaussian mutation. These mechanisms aim to enhance population diversity and improve convergence during the evolutionary search process.

The performance of the proposed algorithm is evaluated using real financial data and compared with the approach proposed by Beasley et al. Experimental results demonstrate that the proposed method achieves improved tracking accuracy and competitive portfolio performance.