Pricing VIX Options with Artificial Neural Networks

Abstract

This thesis will study the use of neural networks as a nonparametric method for the pricing of VIX options. We will be training a network to do the pricing, and compare different network architectural choices to achieve more accurate performances. The performance of the network will be compared to traditional Black-Scholes closed form style pricing models. This paper will expand on the current literature of neural networks in pricing options by examining the effectiveness of different architectures on a novel underlying. Finally, we apply our network to the pricing of VIX options during the COVID-19 pandemic in 2020 to see if the neural network approach outperformed traditional pricing methods in an unprecedented time.