Palmer, Samuel; (2019) Evolutionary Algorithms and Computational Methods for Derivatives Pricing. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Palmer_thesis.pdf Download (5MB) | Preview

Abstract

This work aims to provide novel computational solutions to the problem of derivative pricing. To achieve this, a novel hybrid evolutionary algorithm (EA) based on particle swarm optimisation (PSO) and differential evolution (DE) is introduced and applied, along with various other state-of-the-art variants of PSO and DE, to the problem of calibrating the Heston stochastic volatility model. It is found that state-of-the-art DEs provide excellent calibration performance, and that previous use of rudimentary DEs in the literature undervalued the use of these methods. The use of neural networks with EAs for approximating the solution to derivatives pricing models is next investigated. A set of neural networks are trained from Monte Carlo (MC) simulation data to approximate the closed form solution for European, Asian and American style options. The results are comparable to MC pricing, but with offline evaluation of the price using the neural networks being orders of magnitudes faster and computationally more efficient. Finally, the use of custom hardware for numerical pricing of derivatives is introduced. The solver presented here provides an energy efficient data-flow implementation for pricing derivatives, which has the potential to be incorporated into larger high-speed/low energy trading systems.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item