Study of CNN-Based News-Driven Stock Price Movement Prediction in the A-Share Market

Abstract

Stock market has always been an important research field of scholars in various industries, and short-term stock price forecasting is the focus of both finance and computer research. This paper applies the convolutional neural network (CNN) to short-term stock price movement forecasting by using daily stock news of a famous company called Guizhoumaotai in the Chinese wine industry. Two scenarios were taken into consideration: first, the news occurred in a day’s transaction time was used to predict the day’s stock price movement; second, the news occurred before a day’s opening time and after the transaction time of the previous day was used to predict the day’s stock price movement. In addition, the stock attentions of Baidu search index and/or media index were added into the model to explore whether they have significant improvement on prediction. The experimental results show that using the news data of the day can achieve better prediction performance. In addition, the introduction of Baidu Index improves the result of stock price prediction to some extent however, with a little effect.