Deep Learning based Long Short-Term Memory Recurrent Neural Network for Stock Price Movement Prediction

Abstract

Background/Objectives: Stock price movement prediction is a difficult task that is simulated using machine learning algorithms to anticipate stock returns. Methods: This study uses the Long-Short-Term Memory (LSTM) Recurrent Neural Network deep learning algorithm combined with the stock’s price action method to predict the movement of intraday price for short-term forecasting. The dataset uses data points such as date, open, high, low, close, volume. The predictions of price movement accuracy were tested on State Bank of India (SBI) stock and one year of the trading dataset used for training the algorithm. Findings: The proposed algorithm gives the prediction of price movement accuracy is up to 98.9%, MSE is 0.918 and MAPE 0.987 with one year of the training dataset. The SBI share price can be predicted one day before and the price prediction can be range level, which means upward or downward. The proposed method has proven to be better than traditional machine learning methods in terms of prediction accuracy and speed. Novelty: This research suggested a fine-tuned and personalized deep learning prediction system that coupled the price action technique with LSTM to make predictions. The combination of the price action method with deep learning algorithm in forecasting is not tried before but this paper does. Keywords: Long­Short­Term Memory, Stock Trading, Stock Price prediction, Recurrent Neural Network, Deep Learning, Machine Learning