EAN: Event attention network for stock price trend prediction based on sentimental embedding

It is only natural that events related to a listed company may cause its stock price to move (either up or down), and the trend of the price movement will be very much determined by the public opinions towards such events. With the help of the Internet and advanced natural language processing techniques, it becomes possible to predict the stock trend by analyzing great amount of online textual resources like news from websites and posts on social media. In this paper, we propose an event attention network (EAN) to exploit sentimental event-embedding for stock price trend prediction. Specially, this model combines the merits from both eventdriven prediction and sentiment-driven prediction models, in addition to exploiting sentimental event-embedding. Furthermore, we employ attention mechanism to figure out which event contributes the most to the result or, in another word, which event is the main cause of the price fluctuation. In our model, a convolution neural network (CNN) layer is used to extract salient features from transformed event representations, and the latter are originated from a bi-directional long short-Term memory (BiLSTM) layer. We conduct extensive experiments on a manually collected real-world dataset. Experimental results show that our model performs significantly better in terms of short-Term stock trend prediction.