Multi-objective portfolio optimization using real coded genetic algorithm based support vector machines
Investors need to grasp how liquidity affects both risk and return in order to optimize their portfolio performance. There are three classes of stocks that accommodate those criteria: Liquid, high-yield, and less-risky. Classifying stocks help investors build portfolios that align with their risk profiles and investment goals, in which the model was constructed using the one-versus-one support vector machines method with a radial basis function kernel. This model was trained using a combination of the Kompas100 index and the Indonesian industrial sectors stocks data. Single optimal portfolios were created using the real coded genetic algorithm based on different sets of objectives: Maximizing short-term and long-term returns, maximizing liquidity, and minimizing risk. In conclusion, portfolios with a balance on all these four investment objectives yielded better results compared to those focused on partial objectives. Furthermore, our proposed method for selecting portfolios of top-performing stocks across all criteria outperformed the approach of choosing top stocks based on a single criterion.