A Dual-View Model for Stock Price Prediction of Internet-of-Thing Enterprises

Abstract

In recent years, with the continuous development of the capital market and intelligent internet of things (IIoTs) technologies, investors have focused more on IIoTs enterprises’ stocks. Since the stocks of IIoTs enterprises have the characteristics of heavy capital flows and high stock price volatility, the effective prediction of IIoTs stock price changes plays an extremely important role in improving investment returns and controlling investment risks. According to the above characteristics, our model takes stock trend fluctuations and time series indicator changes into consideration and comprehensively captures IIoTs stock information from both the temporal domain and the spatial domain. Specifically, the proposed model is a dual-view model that incorporates selected trading indicators to predict the closing prices of stocks. In the first view, an RNN model is designed to enlarge the receptive field of the model. In the second view, we introduce an attention mechanism to extract the influences of individual stock trends on the forecasting target. To verify the validity of this prediction method, we compare it with six other stock prediction methods. The results show that on the Ping An (601318) and IFLYTEK (002230) datasets, our method achieves the best results, that is, the lowest RMSE values.