Deep Learning Approaches for Stock Price Prediction: Methods and Trends

Abstract

The ability to anticipate potential opportunities or crises in the stock market has always been highly valued by investors. This skill became particularly crucial during the Covid-19 global pandemic, as effective risk management became essential for navigating the volatile environment. Beyond traditional business analysis strategies, there is a growing demand for a robust intelligent system capable of accurately predicting stock prices to inform investment strategies. Currently, a significant body of research focuses on predicting stock price trends, predominantly employing deep learning techniques. Despite the success of these studies in achieving favorable outcomes, there has been a scarcity of comprehensive surveys summarizing the deep learning methods utilized in stock price prediction.Therefore, this paper aims to provide a thorough review of the machine learning techniques applied in stock price forecasting, outline the development context of this field, and analyze emerging trends based on previously published research. The reviewed papers were categorized according to the deep learning methods they employed, including Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), Recurrent Neural Networks (RNN), and various hybrid deep learning models. Additionally, this paper identifies key datasets, variables, models, and results from each study. The survey presents these findings using widely adopted performance metrics such as Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), and Mean Square Error (MSE).