Quest for efficient option pricing prediction model using machine learning techniques

Abstract

Prediction of option prices has always been a challenging task. Various models have been used in the past but there has been no effort to point out which model is suited best for predicting option prices. Computational time plays an important role in prediction of option prices since these time series are usually large. It is computationally expensive to employ a traditional statistical model which comprises of two phases namely model identification and prediction. A good fitting model may not always be good for prediction due to high fluctuation in the market. Various non parametric models like Multilayer perceptron (MLP), Radial Basis function (RBF) Neural Network and Support Vector regression (SVR) have been employed in the past. MLP and RBF networks take enormous amount of time since the network is learned after a number of iterations. In this paper, prediction of American stock option prices (both call and put options) for companies belonging to various sectors and also prediction of European option prices of Nifty index futures has been attempted using GRNN which has not been attempted so far in the literature. Comparative performance evaluation of GRNN has been done with Support Vector Regression (SVR), MLP and Black Scholes Model. It has been shown that the performance of GRNN is superior to the well known Black Sholes model and other non parametric models like MLP and RBF both in terms of accuracy and time and it performs at par with SVR.