Hedging with Reinforcement Learning

Specialeforsvar: Nick Kjær Mikkelsen

Titel: Hedging with Reinforcement Learning: Optimizing Option Hedging Strategies with Reinforcement Learning in a Market with Transaction Costs

Abstract: In this thesis, we investigate the use of reinforcement learning for hedging options in a market with transaction costs. Simulation was used to model the market, and a reinforcement learning agent was then trained on it and tasked with making the
decisions about when to buy or sell the underlying assets in order to hedge the risk of the options. We define a reward function that takes into account the value of the options and the transaction costs of trading, and we use a reinforcement learning algorithm to train the agent to maximize this reward. However, our results show that the performance of the resulting hedging strategy is only slightly better than traditional delta hedging, and a less frequent delta hedging could achieve the same results.
The thesis presents the theory and knowledge needed to implement a reinforcement learning algorithm in a financial context. We also analyze the limitations of our approach and provide suggestions for future work that could improve the performance of
reinforcement learning for hedging options in a market with transaction costs.

Vejleder: Rolf Poulsen
Censor:    David Sloth Pedersen, Danske Bank