As computers increase their power, machine learning gains an important role in various industries. We consider how to apply this method of analysis and pattern identification to complement extant financial models, specifically option pricing methods. We first prove the discussed model is arbitrage-free to confirm it will yield appropriate results. Next, we apply a neural network algorithm and study its ability to approximate option prices from existing models. The results show great potential for applying machine learning where traditional methods fail. As an example, we study the implied volatility surface of highly liquid stocks using real data, which is computationally intensive, to justify the practical impact of the methods proposed.
Worcester Polytechnic Institute
Major Qualifying Project
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work, subject to other agreements. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted.