With significant improvements in deep learning techniques, computational power and the amount of data available, efficient and effective derivatives pricing is possible. In this paper, we try to apply deep reinforcement learning using recurrent neural network architecture for pricing derivatives. This approach is an enhancement over traditional fair pricing models which are efficient only to price standard derivatives, as they assume a certain underlying distribution for underlying. But Reinforcement learning models can price derivatives directly from the underlying data without assuming any parametric form; hence, certain problems like jump discontinuities and fat-tails that occur in the real world derivatives prices can be avoided using deep reinforcement learning and Recurrent Neural Networks.
mjavedakhtar / deep-reinforcement-learning-for-derivative-pricing-and-trading-strategy Goto Github PK
View Code? Open in Web Editor NEWLicense: MIT License