Probability distortion of truncated quantile critics for stock trading environment

• Main Author: Foo, Marcus Jun Rong
• Other Authors: Patrick Pun Chi Seng
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Business::Finance::Asset allocation; Engineering::Computer science and engineering::Mathematics of computing
• Online Access: https://hdl.handle.net/10356/166624
• Institution: Nanyang Technological University

This paper proposes the use of Cumulative Prospect Theory (CPT) in combination with Truncated Quantile Critics (TQC) for stock trading. CPT is a popular model of decision making under risk that has been shown to better describe human behavior than traditional models such as expected utility theory. TQC is a variant of the popular Quantile Regression DQN algorithm that has been shown to be more sample efficient. By combining these two models, our approach aims to better capture the decision making process of human traders. Furthermore, we incorporate Prelec weighting as a side study to mitigate time inconsistency in decision making. Our experiments in stock trading show that our proposed approach outperforms traditional methods in various portfolio metrics.