O uso de machine learning para previsibilidade do valor de ações da empresa 3M

Abstract

The predictability of stock prices in the financial market represents a constant challenge, consi- dering the complexity and diversity of economic, political, and social factors that influence asset prices. Traditional methods of analysis, although still relevant, often fail to adequately capture the non-linear interactions and complex temporal patterns present in financial data. In light of this, this work proposes and evaluates the effectiveness of a Machine Learning model based on Re- current Neural Networks (RNN), with a Long Short-Term Memory (LSTM) architecture, to more accurately predict the stock value of the 3M company. The solution was implemented with historical stock closing data, using regularization techniques such as Early Stopping and Dropout to improve the model's performance and prevent overfitting. The results obtained were promising, with the model achieving a mean squared error (R2) greater than 70% in the test scenarios, demonstrating the potential of deep learning as a tool for analysis and forecasting in the financial market. Thus, the main contribution of this study is to offer an approach that can assist investors and analysts in making more informed decisions, contributing to the improvement of investment and risk management strategies.