Forecasting the Dubai financial market with a combination of momentum effect with a deep belief network

• Date: 01 July 2019
• DOI: http://doi.org/10.1002/for.2560
• Author: Andreas Karathanasopoulos,Mohammed Osman
• Published date: 01 July 2019

SPECIAL ISSUE ARTICLE

Forecasting the Dubai financial market with a combination

of momentum effect with a deep belief network

Andreas Karathanasopoulos | Mohammed Osman

Dubai Business School, University of

Dubai, Dubai, UAE

Correspondence

Andreas Karathanasopoulos, Dubai

Business School, University of Dubai, PO

Box 14143, Dubai, UAE.

Email: akarathanasopoulos@ud.ac.ae

Funding information

Dubai Financial Market

Abstract

Applying recent advances in machine learning techniques, we propose a

hybrid model to forecast the Dubai financial market general index. Particu-

larly, we exploit a deep belief networks model that applies a restricted

Boltzmann machine as its main component in combination with momentum

effects. We also introduce an innovative way of selecting the inputs by using

momentum effects. With this hybrid methodology we generate a prediction

model along with a comparison of three different linear models. The results

obtained from the hybrid model are better and more stable than the three

linear models. The findings support that the hybrid model we applied will find

their way into finance because of their reliability and good performance.

KEYWORDS

deep belief network, input selection, momentum,transaction cost

1|INTRODUCTION

It is a truism that providing an accurate forecast of the

stock market is one the toughest challenges that both

academicians and non‐academicians encounter. This is

indubitably due to the intrinsic nature of the stock market,

which is characterized by being very complicated,

dynamic, stochastic, chaotic and nonlinear in behavior,

which makes its forecasta very challenging task. Currently,

it is widely recognized that the stock market is one of the

most popular ways to invest and has become an important

means of individual finance in recent times. This essen-

tially entails that investors would liketo estimate the stock

price and select the trading chance accurately in advance,

which could help in bringing higher returns. However,

finding the best time to select this trading chance has

remained a very difficult task for investors because there

are numerous factors that may influence stock prices.

Although having an accurate forecast of the stock mar-

ket is very appealing to both investors andpolicy makers, it

remains an empirical art. Nevertheless, several research

efforts have been carried out to predict the stock market

using different techniques with differing results. With

recent developments in data processing technology,

researchers have culminated their efforts to utilize

machine learning techniques to predict the stock market.

These new machine learning techniques possess the ability

to find patterns, generalize and learn without being explic-

itly programmed. The most popular of these new machine

learning techniques include artificial neural networks,

support vector machines and genetic programming.

In this paper we are proposing a new machine

learning technique called deep beliefs networks (DBN

hereafter) to forecast the Dubai financial market general

index. Karathanasopoulos (2017) was one of the first to

use this methodology in terms of forecasting the crack

spread. He deliberated that this new methodology outper-

forms all the other traditional linear and nonlinear

models. To this end, we are also introducing an innova-

tive way of selecting the inputs to have special features

to be in the machine learning algorithm. Although

reading statistical charts gives a clear picture of the mar-

ket and provides accurate one‐day‐ahead predictions,

nonetheless we are suggesting that the selection of the

Received: 7 June 2017 Revised: 22 May 2018 Accepted: 8 October 2018

DOI: 10.1002/for.2560

346 © 2018 John Wiley & Sons, Ltd. Journal of Forecasting. 2019;38:346–353.wileyonlinelibrary.com/journal/for