Loading...

A Stock Portfolio Management Algorithm Based on Fundamental Market Data for Tehran’s Stock Exchange – Case Study on Mining & Metal Industries

Zarei, Mohammad | 2023

61 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56570 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Habibi, Moslem
  7. Abstract:
  8. The aim of this research is to develop and implement a deep reinforcement learning algorithm for portfolio management in the Tehran stock market, which is considered an emerging market with distinct patterns compared to the stock markets of developed countries. In this study, in addition to the market price data extensively used in previous research, we leverage fundamental ratio data extracted from company financial reports, which have received less attention. Furthermore, the research scope is limited to stocks in the mining and metal industries to enable the utilization of specific industry features, such as susceptibility to global prices of a key commodity. The portfolio management problem in this research is initially formulated as a Markov Decision Process (MDP), and a Double Deep Q-Network (DDQN) reinforcement learning algorithm is employed to solve it. The overall problem-solving process in this research has faced three major challenges: data mismatch between market and fundamental data, infeasible actions, and data scarcity. To address these challenges and improve scalability, the proposed research algorithm utilizes: a) two parallel encoding modules for extracting separate market and fundamental features, b) an optimized mapping function for handling infeasible actions based on a novel method from the literature, c) a minimum or maximum asset weight constraint to enhance diversification, reduce overall portfolio risk, and reduce algorithm execution time, and d) an enhanced search space simulation method with matrix operations. The proposed algorithm is tested on a portfolio composed of three stocks from the Tehran stock market's mining and metal industries and demonstrates a significant cumulative return superiority over the baseline research algorithm and the uniform buy-and-hold strategy
  9. Keywords:
  10. Portfolio Management ; Stock Portfolio Optimization ; Stock Portfolio ; Deep Reinforcement Learning ; Artificial Intelligence ; Parallel Encoder ; Financial Fundamental Data ; Markov Decision Making ; Tehran Stock Exchange

 Digital Object List

 Bookmark

No TOC