Stock market forecasting using recurrent neural network

Abstract

In this research, we study the problem of stock market forecasting using Recurrent Neural Network(RNN) with Long Short-Term Memory (LSTM). The purpose of this research is to examine the feasibility and performance of LSTM in stock market forecasting. We optimize the LSTM model by testing different configurations, i.e., the number of neurons in hidden layers and number of samples in sequence. Instead of using daily stock price data, we collect hourly stock data from the IQFEED database in order to train our model with relatively low noise samples. Nevertheless, based on the prediction results of LSTM model, we build up a stock database with six U.S market stocks from five different industries. The average test accuracy of these six stocks is 54.83%, where the highest accuracy is at 59.5% while the lowest is at 49.75%. We then develop a trade simulator to evaluate the performance of our model by investing the portfolio within a period of 400 hours, the total profit gained by the model is $413,233.33 with $6,000,000 initial investment.