Unveiled by University of Tehran Researchers:
A Novel Model for Portfolio Management in Iran Using Artificial Intelligence
According to a report by the Public Relations Office of the University of Tehran, a recent study by faculty members of the University of Tehran, using deep reinforcement learning, demonstrates how the simultaneous use of price and trading volume data along with technical analysis indicators can significantly improve portfolio returns in Iran's capital market.
This recently published study shows that the use of these advanced algorithms can help investors make far more efficient decisions than traditional methods.
According to the findings of this research, conducted by Dr. Behnam Aminrostamkolaei, a faculty member of the Faculty of Economics, and Dr. Bagher Babaali and Dr. Abolfazl Nadi, faculty members of the College of Science, the designed model—using raw market data (including opening, highest, lowest, and closing prices, and trading volume) alongside indicators such as the Simple and Exponential Moving Averages, Bollinger Bands, the Relative Strength Index (RSI), and the Ease of Movement (EMV) indicator—has been able to record better performance than the market index and classical methods such as equally weighted portfolios and the Markowitz mean-variance approach.
The researchers have pointed to a key finding regarding investment horizons that emerged from this method.
Dr. Babaali, one of the authors of this article, said: "The results show that adding technical indicators to the deep learning model has created a significant improvement in return and risk-adjusted return over one-year and two-year periods, but this improvement was not statistically significant over a three-year horizon, which indicates the model's greater strength in identifying short-term market fluctuations."
He emphasized: "The synergy of data and machine learning is the main reason for the success of this approach—its ability to discover complex patterns hidden in market data."
The faculty member of the College of Science stated: "The architecture used in this model is inspired by successful methods in international financial markets, in which deep neural networks are used for the simultaneous processing of various types of market data and dynamic decision-making. Traditional systems usually lack the ability to adapt to instantaneous market changes, but in this new method, the intelligent agent learns an optimal strategy for asset allocation over time by observing market conditions and receiving rewards (portfolio returns)."
Dr. Babaali added: "This approach can be particularly useful for investors who do not have a high risk tolerance and seek downside risk management. Regarding the prospect of application in the Iranian market, given the complexities and inherent volatility of Iran's capital market, the findings of this study outline a new path for fund managers and professional analysts. The use of reinforcement learning algorithms can not only help optimize asset portfolios but also, by reducing human error and enabling high-speed data analysis, pave the way for more efficient algorithmic trading in the market."
https://www.scilit.com/publications/3ddc8ff7e4a07de2cc4520e5908858c4

Your Comment :