Investigação da capacidade preditiva de Redes Neurais Recorrentes “Long Short-Term Memory” (LSTM) para os preços de fechamento do Mini Índice de 2022 a 2024

Abstract

One of the strategies used by financial market traders and investors to maximize portfolio returns is forecasting future asset prices, either based on historical price analysis or company fundamentals. In this study, we assess the predictive capability of an LSTM neural network in forecasting the closing prices of the Mini Bovespa Index, testing four different models. The first and simplest model considers only the closing price as a variable. The second incorporates additional price-related information. The third includes price data along with a single technical indicator. Finally, the fourth and most complex model combines price data with multiple technical indicators. Comparing these models allows us to evaluate the effectiveness of different variable sets in predicting market behavior.