Option Pricing Using Neural Networks
Date
2019-08-30
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Due to the properties of large transaction volumes, high innovation and positive promotion to the financial market development, options play an essential role. However, the flexible design of options leads to complicated pricing, which makes accurate option pricing a challenging task for a long time. In this paper, two types of neural networks - feed-forward networks such as the radial basis function (RBF), multilayer perceptron (MLP), Modular network and a recurrent deep learning network Long short-term memory (LSTM) - and three stochastic process pricing models (Black-Scholes-Merton model, Heston stochastic volatility model and Merton Jump-diffusion model) are proposed so as to predict European call option prices. Firstly, it generates simulated option prices data from three stochastic process models to test the effectiveness and approximation ability of the neural networks. Secondly, effective factors such as moneyness, time to maturity and greeks via the Black-Scholes-Merton formula are used as input variables for neural networks. Historical data of S&P 500 European option prices empirically analyzes validity and stability of the neural networks. The performance measures R2, statistical error of the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) are used to evaluate the performance of two types of pricing models. It shows that the MLP network with two hidden layers performs best. In addition, neural networks do outperform the pricing ability of stochastic process pricing models.
Description
Keywords
Stochastic process, feed forward neural network, recurrent neural network, option pricing
Citation
Que, D. (2019). Option Pricing Using Neural Networks (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.