Inflexible Hedging in the Presence of Illiquidity and Jump Risks

Market in the real world is inevitably incomplete, and a lot of delicate models under the complete market assumption fails in such a scenario. This paper deals with the hedging problem in incomplete market. It deals with three sources of incompleteness: non-continuous asset prices, illiquidity, and discrete transaction dates. It proposes a jump-diffusion model to describe asset dynamics. Under this model, three neutral network models (RNN, LSTM, Mogrifier-LSTM) with three types of loss functions are implemented and compared. All neutral networks show promising results, and the Mogrifier-LSTM is the fastest model in diverging speed.