Abstract
Forecasting the stock market is one of the most challenging things for investors to do to increase their profits. The stock market is predicted using statistical strategies and learning tools. The objective of this study is to predict the closing price of the stock using Long Short-Term Memory (LSTM) network modified by Sin-Cosine Algorithm (SCA), Autoregressive Integrated Moving Average (ARIMA) and Generalized Auto Regressive Conditional Heteroskedasticity (GARCH) statistical models which is called LSTM-SCA-ARIMA-GARCH model. An evaluation of the proposed method was performed using time series that are of varying stability. In this work, the data of 8 stocks including State Bank of India Network (SBIN), Oracle Corporation (ORCL), Microsoft Corporation (MSFT), Halliburton Company (HAL), Goldman Sachs Group Inc (GS), Cognizant Technology Solution Corporation (CTSH), Bank of America Corp (BAC) and Amazon (AMZN), which included closing stock price have been predicted on a daily and weekly basis, and the daily prediction was more accurate than the weekly prediction. In general, for daily prediction the SCA-LSTM-ARIMA-GARCH model 83.37%, 84.05% and 55.8% better than LSTM, Combination of LSTM and Particle Swarm Algorithm (LSTM-PSO) and LSTM-ARIMA, respectively.