The study fills a void in the DRL hedging field by applying DRL agents to hedge American put options, significantly enhancing existing strategies.
Initial experiments indicate that the DRL agent surpasses the BS Delta and binomial tree hedging strategies, even in the presence of transaction costs.
A unique reward function is introduced, which includes a penalty for transaction costs and the negative absolute difference between the option value and underlying asset position.
Utilizing a stochastic volatility model, the DRL agent not only outperforms the BS Delta strategy in familiar conditions but also shows superior general hedging performance.
#deep-reinforcement-learning #hedging-strategies #american-put-options #stochastic-volatility #financial-models
Collection
[
|
...
]