Document Type : Original Article
Authors
1
PhD. Student in Financial Engineering, Department of Financial Management, Faculty of Management and Economy, Science and Research Branch, Islamic Azad University, Tehran, Iran
2
Assistant Professor, Department of Financial Management, Faculty of Humanities, Shahr-e-Qods Branch, Islamic Azad University, Tehran, Iran and Visiting professor of Science and Research Branch, Islamic Azad University, Tehran, Iran.
3
Assistant Professor, Department of Accounting, Faculty of Economics and Accounting, Central Tehran Branch, Islamic Azad University, Tehran, Iran
4
Assistant Professor, Department of Financial Management, Faculty of Management, Electronic Campus, Islamic Azad University, Tehran, Iran
10.30495/ijfma.2023.69021.1903
Abstract
There has been an increase in the development of automated trading systems in numerous countries, including Iran. The greatest advantage of such systems is that they allow traders to make trading decisions at an increased pace and with more accuracy without having to rely on emotions. Estimating price is one of the most important aspects of algorithmic trading. Deep neural networks are preferred for estimation. Additionally, investors who rely on algorithmic trading have a huge advantage by having a model that estimates opening, maximum, minimum, and closing prices simultaneously. In this study, using short-term LSTM deep-long-term memory neural networks, and the genetic algorithm, these four prices are estimated simultaneously. In addition, the optimal feature was selected by considering 40 price, volume, volumetric and volumetric indicators. The proposed model is evaluated using five shares from the Tehran stock exchange during the period 2012-2021, namely Isfahan oil refining, Iran Khodro, and Amirkabir Kashan Steel, Eqtesad Novin Bank, Chin-Chin Industry and Cultivation, and Exir Pharmacy. Based on the results of this study, the proposed model has excellent simultaneous estimation performance and the average estimation error of all 4 prices is less than 8%, demonstrating that the proposed method has a lower estimation error.
Keywords