Kernel-Based Machine Learning for Option Pricing

Traditional option pricing models such as the Black-Scholes-Merton model have certain limitations, which has given rise to research that asks the option price data itself to reveal which factors drive option prices. In this article, we explore the application of kernel-based learning methods, specifically the Kernel Support Vector Machine (KSVM) and Gaussian Process (GP) models. We illustrate the setup and examine the performance of these two models in explaining call option prices for the S&P 500 index between August 31, 2018, and August 31, 2023. The results for this dataset indicate that the Radial Basis KSVM model is superior to other KSVM models; the Radial Basis KSVM model has a slightly lower mean squared error than the GP model. We find that both models provide reasonably accurate predictions of option prices. This article contributes to the literature on machine learning techniques in finance by highlighting their potential in option pricing.