MStoCast: Multimodal Deep Network for Stock Market Forecast

Stock market analysis is a complex task that involves various types of data, such as web news, historical prices, and technical market indicators. Recent research in this area focuses on analyzing these modalities either separately or all together, but the underlying correlation patterns in the multimodal data were not captured. To address this issue, we propose MStoCast, a Multimodal Stock Market Forecast model that uses innovatively designed deep networks. First, we propose a common network that captures cross-modality and joint information. Then, we construct a unique network that discovers bi-modal information from the inputs. These pieces of information are then integrated and processed through a fully connected layer to predict the direction of the closing price movement. Experiments using real-world datasets show that our MStoCast model significantly outperforms other state-of-the-art models.