An LSTM-GRU based hybrid framework for secured stock price prediction

Abstract The prediction of the stock prices is a very challenging task as the data is associated with nonlinearity and volatility. The machine learning and artificial intelligence methods have been found to make this task more efficient and the advent of high throughput computes have proved to be beneficial in these tasks. In this work a hybrid LSTM-GRU network has been used for prediction of the adjusted closing price of the Standard & Poor 500 index. Also, the initial number of six features have been increased to 25 features by adding several technical indicators. The performance indicators like Return ratio, R2, MSE, Optimism and Pessimism ratios are used to compare the proposed model with stand-alone LSTM, GRU and MLP models. This comparison establishes that the proposed model is capable of more accurate prediction of the stock market prices.