Deep reinforcement learning based trading agents: Risk curiosity driven learning for financial rules-based policy
作者：
Highlights：
• A novel rule-based policy approach is proposed to train automated trading agents.
• A continuous virtual environment was created to boost the training process.
• The risk curiosity-driven learning has progressively improved actions quality.
• The appropriateness and efficiency of self-trained rules are provided.
• The trained agent exhibits transfer learning.
摘要
•A novel rule-based policy approach is proposed to train automated trading agents.•A continuous virtual environment was created to boost the training process.•The risk curiosity-driven learning has progressively improved actions quality.•The appropriateness and efficiency of self-trained rules are provided.•The trained agent exhibits transfer learning.
论文关键词：Deep reinforcement learning,Partially observable Markov decision process,Knowledge uncertainty,Financial data,Trading system,Financial engineering
论文评审过程：Received 23 September 2019, Revised 5 August 2020, Accepted 30 December 2020, Available online 6 January 2021, Version of Record 12 January 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.114553
