Conceptual-temporal graph convolutional neural network model for stock price movement prediction and application

Stock price movement prediction is an important problem for trading decision-making. But it is a challenging task due to the nonlinearity and complexity of the stock trading data. This paper analyzes the linkage effect of price movement among stocks with the same concept segment through the dissipative structure theory, which is one of the major drivers for stock price movement. Considering the time-dimensional and concept-dimensional characteristics of the stock price movement, the stock conceptual-temporal network is constructed and the conceptual-temporal graph convolutional neural network model (CT-GCNN) is designed to map the linkage effect and predict the stock price movement. The experiment is conducted to validate the proposed model by utilizing the Chinese stock trading market data, which shows that CT-GCNN model outperforms the baseline deep learning models. Ten application cases are designed according to the conceptual quantity. The monthly highest stock yield is up to 16.276 and the lowest stock yield is 5.083, which reveals the stability and superiority of CT-GCNN model.