Memetic Crow Search Algorithm and Long Short-Term Memory Network Forecasting System for Stock Prices

The stock market is a complex and constantly changing system. Relevant information such as financial indices, technical indicators, and financial reports can all affect stock prices. Investors adjust their investment strategies to gain profits by analyzing trends of the stock market. Stock market forecasting using the most typical time series data has always been a challenging task for researchers. In recent years, deep learning has been used by many researchers in financial-related fields because of its powerful data processing capability, for tasks such as trend forecasting, trading point forecasting, investment portfolio, and strategy optimization. Although direct application to stock price forecasting is rare, it is critical for investors because they can effectively avoid risks and gain further profits by predicting the closing price on the following day. In view of this, a hybrid crow search algorithm and long short-term memory forecasting system (CSLSTM) for stock price based on the long short-term memory (LSTM) architecture is proposed in this paper. For prediction, this system uses three different types of data and corresponding LSTM neural networks. The data from five stocks and recurrent neural network (RNN), gate recurrent unit (GRU), and LSTM neural network were used in this study. The prediction accuracy of the three types of time series neural networks were compared and evaluated. The test results indicated that the proposed CSLSTM forecasting system outperformed any single model on average, allowing users to profit successfully with simple operations in the final backtesting