ABSTRACT
This chapter covers machine learning methods in option pricing. First, it briefly introduces regression trees, random forests, and neural networks – these methods are advocated as highly flexible universal approximators, capable of recovering highly nonlinear structures in the data. The chapter shows how to implement random forests and deep neural networks with tidy principles using tidymodels or TensorFlow for more complicated network structures. It focuses on so-called supervised learning for regressions, while ML can be specified along a vast array of different branches. The chapter trains different models to learn how to price call options without prior knowledge of the theoretical underpinnings of famous option pricing equation by Black and Scholes. Regression trees are a popular ML approach for incorporating multiway predictor interactions. In Finance, regression trees are gaining popularity, also in context of asset pricing. Random forests address these shortcomings of decision trees. The goal is to improve the predictive performance and reduce instability by averaging multiple decision trees.
