"Application of Deep Feedforward Neural Networks for Predicting Daily Movements in the Peruvian Stock Market"

Abstract

"This study evaluates the effectiveness of a Deep Feedforward Neural Network in predicting the daily movement (Up/Hold/Down), with a binary robustness version (Up vs. Not Up) of stocks in the Lima Stock Exchange. We analyze five highly liquid equities from the Lima Stock Exchange and use three feature groups as inputs: market (prices and volume), technical indicators, and economic variables. The model was trained using the ADAM optimizer and evaluated through classification accuracy. The results show accuracy levels comparable to those reported in several international studies, confirming the feasibility of applying neural networks in the Peruvian market. However, the average performance remains below that observed across a set of 37 reviewed studies, suggesting the need to explore alternative hyperparameters and network architectures. © 2026 IEEE."