MGRU-M1dCNN: A Hybrid Model for Stock Price Prediction

Predicting the stocks price is both intriguing and difficult due to the nonlinear and noisy stock’s data. Stock investors’ investment risk can reduce and their return on investment can be significantly increased by properly predicting the change in stock prices. It is challenging for researchers to forecast stock prices using an individual stock forecasting model. Therefore, in this research we proposes MGRU-M1dCNN, a hybrid model to predict the stock opening price of the next day. This model is consisted of multiple gated recurrent unit (MGRU) and Multiple single dimensional convolutional neural networks (M1dCNN). MGRU employs to extract the valuable feature patterns from opening price of stocks. M1dCNN is used to extract the spatial features of the inputs received from MGRU to predict the stock prices. The outcomes demonstrate that this technique performs best, with the least Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) which are 0.854 and 0.456. When compared to other GRU based models (GRU-CNN, GRU-DNN, and GRU), the MGRU-M1dCNN model is better suited for stock price forecasting and for providing investors a trustworthy means to choose which stocks to buy.