Abstract
Major technological advancements in the 21st century and access to vast amounts
of data have seen the widespread commercial integration of machine learning in
our everyday lives. This was a feat deemed unlikely a few decades ago when these
algorithms were first introduced. The adoption of machine learning techniques in
financial markets has, however, been slow in comparison to other fields. This is
partly due to the availability of sufficient data needed to train and support these
techniques.
The fundamental theory underlying financial derivative pricing has also changed
after the global financial crisis (GFC) of 2008, where it became apparent that certain
assumptions underlying classical derivative pricing theory were outdated and not
reflective of modern financial markets. This led to the development of a multi-curve
framework to address the assumption on the existence of a unique risk-free rate,
which is not necessarily a realistic assumption.
In light of these two major developments, this dissertation investigates how machine
learning techniques such as artificial neural networks (ANNs) can be used to approximate
the prices and option price sensitivities of Johannesburg Stock Exchange
(JSE) Top 40 European call options in both a classical and modern multi-curve
framework. Given the illiquid nature of the South African financial market and the
general lack of sufficient option price data, the ANNs throughout this dissertation
were trained on artificially generated data. The out-of-sample performance of the
ANNs were evaluated using an implied volatility surface constructed from published
volatility skews from the JSE.
The main findings indicate that ANNs trained on artificially generated data are able
to approximate the prices and respective option price sensitivities of both a classical
and modern multi-curve derivative pricing framework in a real-world application to
the South African market with precision.
Publications from this Dissertation
Du Plooy, R. and Venter, P.J. (2021). Pricing Vanilla Options using Artificial Neural
Networks: Application to the South African Market. Cogent Economics and
Finance, 9(1).
Du Plooy, R. and Venter, P.J. (2021). A Comparison of Artificial Neural Networks
and Bootstrap Aggregating Ensembles in a Modern Financial Derivative Pricing
Framework. Journal of Risk and Financial Management, 14(6), 254.
Du Plooy, R. and Venter, P.J. (2024). Approximating Option Greeks in a Classical
and Multi-Curve Framework using Artificial Neural Networks. Journal of Risk and
Financial Management, 17(4), 140.