Modeling of Financial Systems with Neural Networks

Abstract

This study investigates the use of hybrid neural networks for modeling financial systems to address the challenges of market complexity and nonlinear behavior where conventional methods fail. The study uses financial closing price data from three different datasets covering the period from 2010 to 2024. The first dataset includes BBVA, Banorte, Inbursa, the Mexican Stock Exchange (BMV); the second includes ALFA, GAPB, Kimberly-Clark, Inbursa; and the third includes ALSEA, CEMEX, GCC, Grupo CARSO. The hybrid models were built using multilayer perceptrons (MLP), recurrent neural networks (RNN), long term memory networks (LSTM) and transformer architectures in both standard and variational autoencoder configurations. The results show that these networks successfully capture complex patterns, provide accurate predictions, improve generalization and reduce errors, highlighting the potential of hybrid deep learning models for financial time series prediction.