Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 55487 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Manzuri, Mohammad Taghi
- Abstract:
- Throughout history, forming an optimal asset portfolio has been the primary goal of capital owners and managers of investment funds in any economic activity. Achieving this goal is equivalent to trying to minimize the risk caused by the inevitable fluctuations in the capital market and maximizing the overall investment return during the expected period. Investors can operate in various financial markets where there are different stocks and asset classes in each of these markets. The main goal of investors is to identify profitable stocks and form an optimal asset portfolio based on them.Based on this, during the past decades, many studies have been conducted to form and optimize the stock portfolio. In recent years, deep learning methods have shown that they are highly effective for learning complex patterns in big data. Financial time series, including asset prices, asset return rates, and transaction volume, together can form a complex and massive dataset. Deep learning can discover these complex feature vectors from the input data to predict the target output variables, including the rate of return of assets or portfolios, with acceptable accuracy for investors.In this research, while examining related works in the formation and optimization of the asset portfolio, an effort has been made to implement, test, and compare a framework based on the latest achievements of deep learning for the formation of an optimal asset portfolio. For this purpose, we first used deep reinforcement learning to predict the stock market trend. Then we formed an effective feature vector from the predicted trend along with some technical indicators and past prices. We used this feature vector as the input of deep regression models and predicted the daily rate of return of each stock. Finally, using self-organizing maps, we identified uncorrelated stocks and formed our optimal asset portfolio based on the expected return evaluation. Also, this portfolio was adjusted daily for a period of one month. The results of the evaluations show that our proposal reaches a return rate of 40% and a Sharpe ratio equivalent to 3.07 in one month, which is a significant improvement compared to the results of similar studies. Compared to previous studies, the proposed proposal improved the Sharpe ratio by 10.8%.
- Keywords:
- Deep Learning ; Financial Market ; Time Series ; Portfolio ; Portfolio Recommendation
Digital Object List-
محتواي کتاب - view
Bookmark