Designing an Algorithmic Trading System, Based on Deep Learning (Case Study: Tehran Stock Exchange)
محورهای موضوعی : Artificial IntelligenceParia Soleimani 1 , Behzad Soleimani 2 , Mina Bagheriyan 3 , Erfan Taati 4
1 - Department of Industrial Engineering, South Tehran Branch, Islamic Azad University, Tehran, Iran.
2 - Department of Industrial Engineering, South Tehran Branch, Islamic Azad University, Tehran, Iran.
3 - Department of Industrial Engineering, South Tehran Branch, Islamic Azad University, Tehran, Iran.
4 - Department of Industrial Engineering, South Tehran Branch, Islamic Azad University, Tehran, Iran.
کلید واژه: Deep Learning, LSTM-CNN method, Stock market, Tehran Stock Exchange, Trading System,
چکیده مقاله :
With the development of computer systems in recent years, transactions in financial markets have been made available for investors. Artificial intelligence (AI) -based models have also used in the financial markets due to the development of information systems and their ability to store and retrieve the large volumes of financial data. This research presents a new approach to modeling the buying and selling process in the stock market based on deep learning, LSTM, and CNN methods. In the proposed method, the forecast of the future value of stock indicators obtained using the LSTM algorithm is used as the input features of a CNN network. The CNN network as a classification model provides the buy/sell signal for the algorithmic trading system. In addition, the EC-FS model has been used to determine the most appropriate input indicators for the classification model. The proposed model has been evaluated on the Tehran Stock Exchange market and five selected stocks. The implementation results of this model have been compared with other models such as LSTM-MLP, RNN-CNN, and RNN-MLP. As a result, it can be concluded that the LSTM algorithm performs better in forecasting the indicators of selected stocks that will be used as the features of the classification model. It is a general statement that the collaborative LSTM-CNN method is more effective than other methods at training the buying and selling process. This study can aid the stock market of Iran participants for designing the most effective trading strategy.
With the development of computer systems in recent years, transactions in financial markets have been made available for investors. Artificial intelligence (AI) -based models have also used in the financial markets due to the development of information systems and their ability to store and retrieve the large volumes of financial data. This research presents a new approach to modeling the buying and selling process in the stock market based on deep learning, LSTM, and CNN methods. In the proposed method, the forecast of the future value of stock indicators obtained using the LSTM algorithm is used as the input features of a CNN network. The CNN network as a classification model provides the buy/sell signal for the algorithmic trading system. In addition, the EC-FS model has been used to determine the most appropriate input indicators for the classification model. The proposed model has been evaluated on the Tehran Stock Exchange market and five selected stocks. The implementation results of this model have been compared with other models such as LSTM-MLP, RNN-CNN, and RNN-MLP. As a result, it can be concluded that the LSTM algorithm performs better in forecasting the indicators of selected stocks that will be used as the features of the classification model. It is a general statement that the collaborative LSTM-CNN method is more effective than other methods at training the buying and selling process. This study can aid the stock market of Iran participants for designing the most effective trading strategy.
