Global Stock Market Prediction Using Stock Chart Images with Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) Networks Enhanced by Deep Q-Learning

The global financial markets have seen rapid advancements in predictive modelling, particularly in the use of deep learning techniques to forecast stock market trends. This research presents a novel deep learning framework that enhances global stock market prediction by leveraging Convolutional Neural Networks (CNNs) combined with Long Short-Term Memory (LSTM) networks and optimized through Deep Q-Learning to detect patterns across international markets. The model was trained on U.S. stock chart images and tested across 31 countries over a 12-year period. Pre-processing tailored for CNNs and LSTMs enabled it to capture both visual and sequential features. Deep Q-Learning further enhanced decision-making accuracy, leading to consistent returns across diverse markets, including smaller ones. The results show that this integrated approach surpasses traditional methods, offering robust predictions and valuable insights into global market trends, benefiting investors and analysts worldwide.