Algoritma Genetika Hibrid pada Permasalahan Optimasi Portofolio Dinamis

Abstract

Dynamic portfolio optimization is a crucial problem in evolving financial markets, requiring investment decisions to be adjusted over time. Traditional methods such as Linear Programming (LP) and Quadratic Programming (QP) have significant limitations, including dependence on linearity and normality assumptions, input sensitivity, computational inefficiency for large scales, as well as difficulty handling realistic constraints. Genetic Algorithms (GA) offer promising solutions for large solution spaces and non-linear functions, but often face the problem of slow conver- gence or getting stuck on local suboptimal solutions in dynamic optimization. To overcome this, this study proposes a Hybrid Genetic Algorithm (HGA) that integra- tes GA with Hill Climbing local search method. HGA is designed to utilize the global exploration power of GA and the local exploitation power of hill climbing, synergis- tically improving the quality of the final solution and overcoming the weaknesses of both traditional and standard GA methods. Simulation results show that HGA signi- ficantly improves performance over standard GA in dynamic portfolio optimization, especially in terms of solution quality and convergence speed, although it does not always excel in the aspect of robustness to change. Thus, HGA is proven to be better at solving dynamic portfolio optimization problems.