期刊名称：IAENG International Journal of Computer Science
印刷版ISSN：1819-656X
电子版ISSN：1819-9224
出版年度：2021
卷号：48
期号：4
语种：English
出版社：IAENG - International Association of Engineers
摘要：The generation of the stock timing strategy is a crucial task for stock trading. Vast previous studies focus on proposing an end-to-end stock timing strategy based on reinforcement learning. However, it is difficult to explain why the model works, and it takes courage to believe in a black box in real stock trading. In this paper, we propose PPO Enhancement Strategy to modify the trading signal of the base stock trading strategy instead of predicting the direction of stock price directly. The base strategy can be formed by technical analysis, fundamental analysis, and other interpretable models, so that it can increase the interpretability of the trading model. In order to make the result of the PPO Enhancement Strategy robust, we perform extensive experiments on two market Indices and four stocks from American stock markets. The proposed PPO Enhancement Strategy outperforms the benchmarks, the Buyand- Hold Strategy and the Moving Average Strategy, in terms of different evaluation criteria.
关键词：reinforcement learning;enhancing stock trading strategy;moving average strategy;proximal policy optimization;stock trading
