Deep Reinforcement Learning

Hello everyone,



I wish to use the DRL model in a hedging way using futures. Did anybody already work about it?

How to cover 100% of a portfolio or only 30% for example? Could the model answer 100% or 30% by itself?



Best Regards

Laurent

Hi Laurent,



Your idea can be implemented in a DRL model. The "action" can be modified to buy 10%, 30%, or 100% of your capital. Accordingly for the buy logic, you can buy the percentage of capital you intend to use for hedging.



Hope this helps.



 

Hi Rekhit,



Thanks very much for answer.



Could the model return the percentage of the portfolio to be covered?

How could this be implemented?



Regards

Laurent 

Hi Laurent,



It is possible, in principle. However, it would require a different approach than the trading example from the course on deep reinforcement learning. Input features would consist of volatility, liquidity and exposure metrics, actions would be hedging levels. The reward function needs careful consideration and is probably some kind of risk metric that depends on the specific use case.