Previsão de preços de ações utilizando inteligência artificial

Abstract

The use of artificial intelligence can, through time series data analysis, predict the trend of stock prices and help decision making. This paper aimed to analyze and apply data analysis techniques using artificial intelligence to predict trends in stock prices and help investors about which decision to make when buying or selling. Three machine learning algorithms were used: ARIMA, PROPHET and LSTM, where all the tests were performed in Google Colab. After defining the best model, a web application called The Seer was developed using the Django framework, which allowed for easier forecasting. The models were validated based on the daily closing values of two stocks and one ETF traded on the BM&F Bovespa. Both models obtained better results for periods of up to 60 days, and the ARIMA and PROPHET models obtained better accuracy for periods of up to 30 days. The LSTM model obtained better RMSE and MAPE metrics for the 90 days observed. Finally, it can be concluded that the work achieved the proposed objective