Stock Price Prediction Using LSTM Model in Machine Learning

Abstract

The stock market has been popular in demand since the beginning of industrialization. Different investors and firms related to it find patterns that help in understanding predicting market trends. Within the realm of stock market prediction methods, two primary categories exist: traditional and computerized methods like artificial intelligence (AI). Statistic approaches encompass techniques like the logical regression model with ARCH model, whereas AI methods involve the adoption of different machine learning techniques like the multi-functional perceptron, CNN, naive Bayes network, backpropagation network, single-layer LSTM, SVM, and RNN. Nonetheless, much of such research solely focuses on predicting a single value. To address the need for predicting multiple values within a single model, the development of a model capable of handling different input and simultaneously processes associated output value has been proposed. This model is based on a deep recurrent neural network incorporating the long short-term memory network. Through this approach, it can predict the open price, lowest price, and highest price of a stock concurrently. The connected network model, the LSTM network model, and the deep recurrent neural network model were compared. The connected model beat the other models in the trial results, accurately predicting several variables at once with a precision of more than 97%.