VOL. 10, NO. 23, DECEMBER 2015 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.
www.arpnjournals.com
17486
PRICE PREDICTIVE ANALYSIS MECHANISM UTILIZING GREY WOLF
OPTIMIZER-LEAST SQUARES SUPPORT VECTOR MACHINES
Zuriani Mustaffa
1
and Mohd Herwan Sulaiman
2
1
Faculty of Computer Systems & Software Engineering, Universiti Malaysia Pahang, Lebuhraya Tun Razak, Gambang, Kuantan,
Pahang, Malaysia
2
Sustainable Energy & Power Electronics Research Cluster, Fakulti Kejuruteraan Elektrik & Elektronik, Universiti Malaysia Pahang,
Pekan, Pahang, Malaysia
E-Mail: zuriani@ump.edu.my
ABSTRACT
A good selection of Least Squares Support Vector Machines (LSSVM) hyper-parameters' value is crucial in order
to obtain a promising generalization on the unseen data. Any inappropriate value set to the hyper parameters would directly
demote the prediction performance of LSSVM. In this regard, this study proposes a hybridization of LSSVM with a new
Swarm Intelligence (SI) algorithm namely, Grey Wolf Optimizer (GWO). With such hybridization, the hyper-parameters
of interest are automatically optimized by the GWO. The performance of GWO-LSSVM is realized in predictive analysis
of gold price and measured based on two indices viz. Mean Absolute Percentage Error (MAPE) and Root Mean Square
Error (RMSPE). Findings of the study suggested that the GWO-LSSVM possess lower prediction error rate as compared to
three comparable algorithms which includes hybridization models of LSSVM and Evolutionary Computation (EC)
algorithms.
Keywords: gold price predictive analysis, grey wolf optimizer, least square support vector machines.
INTRODUCTION
The unpredictability of gold price in recent years
has attracted much attention from many parties which
includes commodities traders, mining companies,
investors and academia as well. Govern by high
nonlinearity features, the price of gold has experienced an
expeditious increases during the last several years [1] and
is continually predicted to be on steady condition in 2015
[2]. Nonetheless, to accurately predict the price of gold is
such a great challenge. With the uncertainties of world
economic and surrounded by various factors, this
challenge has paved a positive way for academic
community in exploring a new method for predictive
analysis purposes.
In literature, the application of well-known
machine learning algorithm, namely Artificial Neural
Network (ANN) has been proposed for the said task and
this includes for gold price [3, 4]. Nonetheless, the
adaptation of Empirical Risk Minimization (ERM) which
tends to minimize the training error rather than the true
error makes the ANN suffer with over fitting problem [5].
Recently, the emergence of Statistical Learning Theory
based algorithm, viz. Least Squares Support Vector
Machines (LSSVM) [6] has been a rival to the ANN. As
opposed to ANN, LSSVM which is a modification version
of Support Vector Machines (SVM) [7] adopted Structural
Risk Minimization (SRM) principle which minimizes the
generalization error instead of training error [8]. Hence,
promising generalization can be obtained.
Owing to its remarkable nonlinear mapping
capabilities, LSSVM has been proven to contribute a
significant impact in solving various data mining tasks
which includes prediction, classification and many others
[9, 10]. However, despite its diversity in application, it is
well documented that the generalization performance of
LSSVM is highly dependent on the value of its two free
hyper-parameters, namely regularization parameter, γ and
kernel parameter, σ2 [11]. Any improper value set to the
hyper parameters would result in undesired generalization
performance.
An extensive literature reviews reveals that a
good numbers of meta-heuristic algorithms have been
proposed in order to cater this matter. In [11-13], the
LSSVM is hybrid with a Swarm Intelligence (SI)
algorithm, namely Particle Swarm Optimization (PSO) for
parameter tuning. In the studies, the efficiency of PSO-
LSSVM is realized in different problem domain which
includes nuclear science, shipping and water drainage and
irrigation respectively. On the other hand, the capability of
Genetic Algorithm (GA), which is a dominant algorithm in
Evolutionary Algorithm (EA) class was tested in several
function estimation problems, which includes in [14, 15].
Meanwhile, in [16], the LSSVM is hybridized with Fruit
Fly Optimization (FFO) [17] for electric load predictive
analysis. In the study, the FFO which is inspired from the
food searching behaviour is employed as an optimizer to
LSSVM. Later, the FFO-LSSVM is compared against
several identified techniques which include single LSSVM
and regression technique. Final results suggested that the
FFO-LSSVM is capable to produce lower error rate
relative to several identified metrics.
With respect to that matter, in this study, the
LSSVM is optimized utilizing a new SI algorithm, namely
Grey Wolf Optimizer (GWO) [18] which was originally
inspired by the collective behaviour of grey wolf. This
algorithm has been proven to be competitive to the
existing meta-heuristic algorithms which include the PSO,
GSA, Differential Evolution (DE), Evolutionary
Programming (EP) and Evolution Strategy (ES) [18]. With
such an impressive performance, in this study, the GWO is