Multi - Modal Deep Learning Model for Stock Crises

This research proposes a novel model combining LSTM, ResNet, and attention mechanisms for early warning signals of stock crises. The model integrates stock return time series data, economic indicator images, and attention mechanisms to predict stock crises. Experiments using Nasdaq 100 stock data demonstrate promising performance compared to baseline models. The study highlights the potential of deep learning and attention mechanisms in stock market prediction. Limitations and future directions are discussed. This research contributes to the field of stock market prediction and crisis detection.