Hybrid Information Mixing Module for Stock Movement Prediction

With the continuing active research on deep learning, stock price prediction using learning has been actively conducted in financial industry. This paper proposes a method for predicting movement and news data. The market is affected by many variables; thus, volatility should be considered movement. Because markets are efficient, all kinds of information quickly reflected prices. We create new fusion mix combining text data features propose hybrid mixing module designed two map blocks effective interaction between features. extract multimodal time-series semantic In this paper, multilayer perceptron-based model, module, applied to conduct fluctuation experiment with high volatility. addition, accuracy, Matthews correlation coefficient (MCC) F1 score were used verify performance module.