Loading...

A Combined Sentiment Analysis and Deep Reinforcement Learning Approach in Portfolio Formation

Beydaghi, Elham | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 58301 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Manzuri, Mohammad Taghi
  7. Abstract:
  8. Forecasting the market trend has been one of the most challenging and hot topics among investors and shareholders. In fact, predicting stock trends leads to better choices for forming a suitable and optimal asset portfolio. To achieve this goal, it is necessary to try to minimize the risk caused by the inevitable fluctuations in the capital market and to maximize the overall investment return during the expected period. Forecasting the stock market can be very complicated due to the influence of stock prices on many factors, such as financial and monetary markets, news and political-economic-social events. In recent years, deep reinforcement learning methods have proven to perform well in learning complex patterns, such as financial time series. To model this huge data set, it is not possible to just rely on the ideas of financial economics. On the other hand, one of the important issues in stock portfolio management is adding the excitement factor and its psychological effect on capital owners' decisions in the stock market. As a result, a tool can be effective in the process of forecasting the stock market, which can find the feature vector by analyzing the sentiments of the stock market as well as the financial time series input data and then use it to predict the return rate of the asset. Therefore, in this research, we seek to simulate and evaluate the automatic trading process for the optimal formation of the stock portfolio by using technical indicators as well as market sentiment analysis. In this research, while examining related works in forecasting the stock market and forming an asset portfolio, we proposed an approach in which we were able to predict the stock market with a mean square error of 0.01 and also build an asset portfolio with a Sharpe Ratio of 1.87
  9. Keywords:
  10. Deep Reinforcement Learning ; Financial Market ; Time Series ; Market Sentiment ; Sentiment Analysis ; Portfolio Recommendation

 Digital Object List

 Bookmark

No TOC