Acknowledgements

This study stems from part of my PhD thesis (download here) “Empirical Assessment and Model Selection in Option Pricing”.

An early work can be found as a technical report at SSRN. Name of the publication is Cluster Stability of Error Representation in Option Pricing. The main difference between early work and this one is the early work claims that data mining algorithms can yield better performance than arbitrarily defined boundaries. This work is more focused on finding the best algorithm with the parametrization that brings the highest cluster prediction stability. We aim to increase the number of models and variety of parametrizations (It is still a work in progress, though…). As a trade-off, this work will not include as many assets, pricing error types and pricing models as the early work (Still technically doable with learnfin).

This work is in collaboration with Prof. Refik Güllü (my PhD supervisor) and Assoc. Prof. Wolfgang Hörmann. Early work (i.e. PhD Thesis and technical reports) is supported by Bogazici University Scientific Research Projects (Project number 8101).

Updated: