强化学习在风险规避型数字化资产配置中的应用

发布时间：2018-06-13 19:29

本文选题：数字化资产配置 + 强化学习　；参考：《山东大学》2017年硕士论文

【摘要】：近年来,强化学习在数字化资产配置和投资组合管理中应用日益加深。但是,绝大部分强化学习算法在应用于资产配置的过程中并不考虑风险这一因素,从而给投资者带来的收益具有较强的波动性。这种风险中性的资产配置决策带来的高波动性往往给投资者带来较大的资产回撤。在资本市场中,无论是机构投资者还是个人投资者,在面临较大的资产回撤时,往往选择中断投资并退出市场。这种做法规避进一步损失的同时也丧失了继续盈利的机会,是投资者十分希望避免的局面。本文在传统的强化学习模型引入风险这一因素,通过对风险项——资产最大回撤进行效用惩罚,提出了风险规避的强化学习模型。
[Abstract]:In recent years, reinforcement learning has been increasingly applied in digital asset allocation and portfolio management. However, most reinforcement learning algorithms do not take risk into account in the process of asset allocation, so the return to investors has strong volatility. The high volatility of this risk-neutral asset allocation decision often brings about a larger asset retreat for investors. In capital markets, both institutional and individual investors often choose to interrupt their investments and withdraw from the market when they are faced with larger asset recoveries. This approach avoids further losses while also losing the opportunity to continue to make profits, a situation that investors are keen to avoid. This paper introduces the risk into the traditional reinforcement learning model, and puts forward the reinforcement learning model of risk aversion by punishing the maximum withdrawal of the risk term.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51

【相似文献】