Design of an investment portfolio through Machine Learning and Swarm Intelligence

Resumen

An investment portfolio is a collection of financial instruments where an investor allocates certain amounts of money to obtain a maximum profit with a minimum risk, modeled by Markowitz's Optimal Portfolio Theory, which is why the selection of financial assets and the allocation of amounts to invest are key. In this proposal, a single-objective optimization problem is proposed with the aim of maximizing the profit/risk ratio. The selection in the proposal documented in this paper is carried out by means of a clustering algorithm known as kmeans on the assets of S&P/BMV IPC of the Mexican Stock Exchange. On the other hand, the allocation of investment amounts is made after comparing the classical analytical method of the Generalized Reduced Gradient with two softcomputing methods: Particle Swarm Optimization, and the Whale Optimization Algorithm, a heuristic method that is based on whale hunting activities. Both methods are analyzed by means of statistics to determine the stability of both with respect to their executions. The results obtained in this work show that the best alternative studied is the Whale Optimization Algorithm, since it has a faster convergence and a higher objective function value in relation to the swarm of particles.