An Adaptive Neuro-Based Fuzzy Inference System (ANFIS) for the Prediction of Option Price: The Case of the Australian Option Market

An Adaptive Neuro-Based Fuzzy Inference System (ANFIS) for the Prediction of Option Price: The Case of the Australian Option Market

Hooman Abdollahi (System Dynamics Group, University of Bergen, Norway)
Copyright: © 2020 |Pages: 19
DOI: 10.4018/IJAMC.2020040105

Abstract

Option price prediction has been an important issue in the finance literature within recent years. Affected by numerous factors, option price forecasting remains a challenging problem. In this study, a novel hybrid model for forecasting option price consisting of parametric and non-parametric methods is presented. This method is composed of three stages. First, the conventional option pricing methods such as Binomial Tree, Monte Carlo, and Finite Difference are used to primarily calculate the option prices. Next, the author employs an Adaptive Neuro-Fuzzy Inference System (ANFIS) in which the parameters are trained with particle swarm optimization to minimize the prediction errors associated with parametric methods. To select the best input data for the ANFIS structure, which has high mutual information associated with the future option price, the proposed method uses an entropy approach. Experimental examples with data from the Australian options market demonstrate the effectivity of the proposed hybrid model in enhancing the prediction accuracy compared to another method.
Article Preview
Top

Introduction

Option price is the amount that the buyer of an option contract is charged to acquire the right to buy or sell a security at a specified price in the future. Depending on the market status, option holders have the choice whether or not to exercise the options. Option price is used by investors to manage investment risks. Therefore, a reliable price forecasting is very valuable for investors in order to make a wise decision.

Option price in a market is volatile and affected by several factors such as the asset price, interest rate, exercise price, expiration time, etc. Hence, the option price forecasting is an intricate problem. The moment of predicting the option price along with complications rooted in its process stimulated numerous researchers. Several parametric (e.g. Cox et al., 1979; Tian, 1993; Heston, 1993; Hutchinson et al., 1994; Bakshi et al., 1997; Hull, 2006) and non-parametric methods (e.g. Lajbcygier and Jerome, 1997; Cao and Tay, 2001; Keber and Schuster, 2001; Schittenkopf and Dorffner, 2001; Tsitsiklis, 2001; Zapart, 2003; Wang, 2009; Xi et al., 2014) have been introduced for this aim.

The common parametric approaches are the binomial tree (BT), finite difference (FD), and Monte Carlo (MC) simulation methods (Hull, 2006). In the recent years, non-parametric methods including neural network (NN) (Amilon, 2003; Binner et al., 2005; Liang et al., 2009; Lin and Yeh, 2009), fuzzy neural network (FNN) (Leu et al., 2010; Yu et al., 2011), and support vector regression (SVR) (Cao and Tay, 2001; Liang et al., 2009) are also used in the option price forecasting. The NNs and FNNs provide a platform for modeling nonlinear input-output mapping functions. The proper choice of input features is a pivotal element for the success of the NN or FNN for option price forecasting models. Moreover, the accurate tuning of adjustable parameters secures the efficiency of these models.

There are many approaches to be employed in order to accost forecasting problems (Abdollahi and Ebrahimi, 2019). Both parametric and non-parametric methods have been applied in previous works. Various parametric models are documented in researches conducted by Carr et al. (2003), Madan et al. (1998), and Merton (1976). Moreover, recent advancements in the IT technology have enhanced the application of non-parametric models and meta-heuristic algorithms in the pertinent literature in order to predict the option price in different markets. For instance, Gencay and Qi (2001) suggested a NN model with Bayesian regulation to forecast the price of options for a three-month horizon using data from the S&P 500 index call options over the period 1988-1993. Also, Gradojevic et al. (2009) tried to price the S&P 500 European call options using a modular NN from January 1987 to October 1994. Han and Lee (2008) applied a Gaussian processes regression to forecast day-ahead option price for KOSPI 200 ELW options. Later, Park and Lee (2012) proposed a Gaussian processes regression with a positive transformation measure to forecast the distribution of call option values using the KOSPI 200 index option prices over the period from 2008 to 2010. Investigating the performance of parametric and non-parametric methods on the KOSPI 200 index options from 2001 to 2010, Park et al. (2014) proved that non-parametric methods show a better forecasting performance. Table 1 presents a summary of the different methods proposed in recent empirical studies for pricing options.

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 11: 4 Issues (2020): 3 Released, 1 Forthcoming
Volume 10: 4 Issues (2019)
Volume 9: 4 Issues (2018)
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing