Stock Market Prediction Based on Advanced LSTM Models

Deep learning models prove their success in solving time series problems in many domains such as the stock market. In this contribution, we compared deep learning techniques to find the best one with better performance. Four variations of the model Long short-term memory (LSTM) were chosen, and the comparison was made based on the benchmark Mean Square Error (MSE) of the test data. In addition, another Comparison was made between the LSTM variations and two other state-of-the-art models, namely the Simple Recurrent Neural Network and Gated Recurrent Unit. The results of all simulations proved that LSTM variations perform better with the lowest MSE. Finally, a solution to forecast the next day’s closing price was proposed using deep learning models.