International Journal of Quality & Reliability Management

An integrated system based on fuzzy genetic algorithm and neural networks for stock
price forecasting: Case study of price index of Tehran Stock Exchange
Noraddin Mousazadeh Abbassi Mohammad Ali Aghaei Mahdi Moradzadeh Fard
Article information:
To cite this document:
Noraddin Mousazadeh Abbassi Mohammad Ali Aghaei Mahdi Moradzadeh Fard , (2014),"An integrated
system based on fuzzy genetic algorithm and neural networks for stock price forecasting", International
Journal of Quality & Reliability Management, Vol. 31 Iss 3 pp. 281 - 292
Permanent link to this document:
http://dx.doi.org/10.1108/IJQRM-06-2012-0085
Downloaded on: 18 October 2016, At: 18:36 (PT)
References: this document contains references to 25 other documents.
Downloaded by UFRB At 18:36 18 October 2016 (PT)

To copy this document: permissions@emeraldinsight.com

The fulltext of this document has been downloaded 386 times since 2014*
Users who downloaded this article also downloaded:
(2010),"Performing stock price prediction use of hybrid model", Chinese Management Studies, Vol. 4 Iss 1
pp. 77-86 http://dx.doi.org/10.1108/17506141011033016
(2000),"Artificial Neural Network and the Financial Markets: A Survey", Managerial Finance, Vol. 26 Iss 12
pp. 32-45 http://dx.doi.org/10.1108/03074350010767034

Access to this document was granted through an Emerald subscription provided by emerald-srm:571046 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.

*Related content and download information correct at time of download.

 The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0265-671X.htm

QUALITY PAPER Fuzzy genetic

algorithm and
An integrated system based neural networks
on fuzzy genetic algorithm and
281
neural networks for stock
price forecasting Received 13 June 2012
Revised 9 June 2013
Accepted 14 June 2013
Case study of price index of Tehran Stock
Exchange
Noraddin Mousazadeh Abbassi
Tarbiat Modares University, Tehran, Iran
Downloaded by UFRB At 18:36 18 October 2016 (PT)

Mohammad Ali Aghaei

Accounting Department, Tarbiat Modares University, Tehran, Iran, and
Mahdi Moradzadeh Fard
Accounting Department, Azad Islamic University Branch Karaj, Karaj, Iran

Abstract
Purpose – The aim of this research is to predict the total stock market index of the Tehran Stock
Exchange, using the compound method of fuzzy genetics and neural network, in order for the active
participants of the finance market as well as macro decision makers to be able to predict the market trend.
Design/methodology/approach – First, the prediction was done by neural network, then the
output weight of optimum neural network was taken as standard to repeat this prediction using the
genetic algorithm, and then the extracted pattern from the neural network was stated through
discernible rules using fuzzy theory.
Findings – The main attention of this paper is investors and traders to achieve a method for
predicting the stock market. Concerning the results of previous research, which confirms the relative
superiority of non-linear models in price index prediction, an appropriate model has been offered in
this research by compounding the non-linear method such as fuzzy genetics and neural network. The
results indicate superiority of the designed system in predicting price index of the Tehran Stock
Exchange.
Originality/value – This paper states its originality and value by compounding the non-linear
method issues pattern to predict stock market, to encourage further investigation by academics and
practitioners in the field.
Keywords Artificial neural network, Fuzzy genetic system, Fuzzy theory, Genetics algorithm
Paper type Case study

Introduction International Journal of Quality &

Nowadays, those who win are those who are able to predict the future, or at least some Reliability Management
Vol. 31 No. 3, 2014
future information, upon which they can base their decisions. pp. 281-292
To predict what will happen in the future, often we rely on the information provided q Emerald Group Publishing Limited
0265-671X
by historical events, in such a way that past data is analyzed in order to produce an DOI 10.1108/IJQRM-06-2012-0085
 IJQRM applicable model for future trends. In most of the prediction methods, it is supposed
31,3 that the relationship between variables will be going into the future.
Determining stock market tendencies is one of the most challenging issues in market
debates, to the extent that there are so many elements affecting the function of the stock
market. A few instances of these are political events, the general economic environment
and also the expectations of traders (Tsang et al., 2007; Hadavandi et al., 2010).
282 Investors use various and often smart systems in the stock market to make
decisions about future trades in order to obtain an optimum output. However, financial
experts believe that, since the market trend is non-linear, non-certain and dynamic,
access to a precise prediction is so difficult that it makes it impossible for the experts to
agree on any one of the current prediction methods of financial time series as the most
effective method (Tsang et al., 2007).
Artificial neural networks are one of the prediction methods available in this subject
which are more effective in comparison with others (Zadeh, 1994), so that various studies
indicate that artificial neural networks are more precise in prediction than regression
models, and they also offer distinct analysis (Refenes et al., 1994; Yoon et al., 1993).
Downloaded by UFRB At 18:36 18 October 2016 (PT)

An artificial neural network is an idea for data processing whose inspiration comes
from the biological neural system and which processes data like a human brain. This
characteristic makes the neural network come up with data patterns which are hard for
humans to understand. However, since the neural networks are similar to continuous
black boxes, it makes perceiving the solution to a problem difficult with these nets
(Mantas et al., 2006). Researchers believe that the problem can be solved by extracting
the produced rule which represents the patterns extracted by the net.
A genetic algorithm search for hidden patterns in the data is inspired by evolution
theory, in order to satisfy the existing standard. To do so, first a number of accidental
responses would be produced for the given problem, and these primary responses
would be evolved in the next steps using evolution genetic principles, and then
converted into appropriate responses.
Since it is impossible to model mathematically with the more complicated systems,
and the precision of the mathematical model will also be decreased, fuzzy computations
were suggested for such a complicated problem. In fuzzy logic, a function is ascribed to
each variable rather than a number in its space, which is representative of the amount a
variable belongs to any point of space.
Since decision making is always carried out with uncertainty, we are seeking to
model uncertainty in decisions related to value investigating, and to offer a method
compounding fuzzy theory and genetic algorithm in order to predict the price index of
the Tehran Stock Exchange (TEPIX).

Literature review
Tsang et al. (2007) offered a model for the prediction of stock prices, using a compound
of wavelet transform, recurrent neural network and bee colony algorithm. First, they
disintegrated the price time series using Haar wavelet, then the prediction was done by
recurrent neural network, and the obtained weights of neural network were optimized
by bee colony algorithm. The offered model was examined on data from the Dow Jones
Industrial Average (DJIA), the FTSE 100 index, the London Stock Exchange (FTSE),
Nikkei 225, the Tokyo Stock Exchange (Nikkei) and the Taiex index, Taiwan Stock
Exchange. The given model was compared to a compound model of neural network
 and bee colony algorithm, fuzzy time series and fuzzy neural network (ANFIS). Fuzzy genetic
The suggested model had fewer errors than the other models in the all examined cases. algorithm and
Hadavandi et al. (2010) offered a model for the prediction of stock prices, using a
compound of neural network and fuzzy genetic. They used the mentioned model to neural networks
examine information gathered from IT and airline industries on the New York Stock
Exchange. The suggested model was compared to ARIMA and genetic algorithm and
neural network, which were used in prediction, and in all cases it gave a better result 283
than the previous models.
Chang and Liu (2008) used the fuzzy rule of TSK type to predict stock price. TSK
fuzzy model considers technical index as input variables and the obtained result is a
linear compound of input variables. This model is examined using data from the
electronic Stock Exchange Corporation of Taiwan and the obtained results indicate a
precision close to 97.6 percent in the TSE index and 98.08 percent in Media Tek.
Since in the stock market intellectual investors revise their predictions according to
the most recent prediction errors, Cheng et al. (2008) used a new time series model to
Downloaded by UFRB At 18:36 18 October 2016 (PT)

decrease the prediction error in the Taiwan stock market. The results indicate the
superiority of this model to the Chen and Yu model.
Chen et al. (2008) used a fuzzy time series model for short-term predictions of the
Taiwan and Hong Kong stock market price. The experimental results obtained from
this research indicate the fact that the traditional statistical method, and the offered
model, both make it clear that stock price patterns in these two markets are short-term.
Chu et al. (2009) used the two element fuzzy time series model for stock index
prediction. The stock index and the amount of trade are considered (in this article) as
elements which are effective in price index predictions. The results indicate this
model’s capability in stock index prediction.
Lin et al. (2007) used the genetic algorithm to make stock market predictions.
The significant factor in a trading rule success is the selection of degrees for all
parameters and their combinations. However, the range of parameters changes in a
large area and the problem is to find the optimum parameter combinations. Genetic
algorithm is used in this article to solve the problem.
Chen et al. (2007) used fuzzy time series based on the Fibonacci sequence to predict
stock price. A time period of five years of data for TSMC, and a time period of 13 years
for TAIEX, was used in this research. The obtained model is superior to the prevalent
fuzzy time series model.
Tsang et al. (2007) analyzed the usage of NN5 in stock price prediction in Hong Kong.
This system has been examined on stock data of two banking stock corporations of Hong
Kong and Shanghai. The system indicates a total success rate of more than 70 percent.
Salcedo-Sanz et al. (2005) indicate the usage of genetic planning in predicting
bankruptcy of companies with no life insurance. Genetic planning has been compared
to other classifying methods in this article.
Farnsworth et al. (2004) used genetic planning to predict a daily output of S&P 500
index amount which indicates the hypothesis of market efficiency. S&P 500 is one of
the most common indexes which has been studied worldwide.
Refenes et al. (1994) have compared the function of stock price treatment model with
regression models through its modeling by neural network. Neural networks are used
as a substitution for classical statistic techniques in this research and these networks
 IJQRM have been used to predict large companies’ stocks. The results indicate that neural
31,3 networks function better than statistic techniques and they present superior models.
A review of existing literature indicates that traditional predicting methods have been
mostly used in the Tehran Stock Exchange and other areas in the world. Concerning the
fact that, to use a traditional pattern, one should use static time series and since most
economic time series are non-static, the traditional patterns have a large amount of
284 trouble when it comes to prediction. Moreover, the artificial intelligence methods do not
have the problem of traditional methods concerning static time series. Regarding these
issues, one of the capable artificial intelligence algorithms, called genetic algorithm, has
been combined to neural network and also fuzzy concepts in this research in order to offer
a predicting method for the price index of the Tehran Stock Exchange (TEPIX).

Genetic algorithm
Inspired by evolution theory and heredity and genetic principles, genetic algorithm
seeks to find an appropriate solution for problems (Goldberg, 1989). To do so, first
some accidental responses are produced for the given problem, and these primary
Downloaded by UFRB At 18:36 18 October 2016 (PT)

responses are then evolved using genetic principles and are converted to the
appropriate response.
In the following part, genetics algorithm steps will be explained (Haupt and Haupt,
1980):
(1) First generation production: the production of Npop accidental responses to the
problem, each of which are appropriately codified in the form of a chromosome.
(2) Valuation: the criterion function determines the value of each chromosome with
respect to their success in solving the problem. The best chromosome of each
generation is called the elite chromosome.
(3) The chromosomes are ordered descending according to criterion function,
only X percent of chromosomes are preserves and the rest will be thrown away. The
possibility of a chromosome’s conservation depends on its merit in such a way that
those chromosomes which are worth more have one more possibility of survival. To
substitute the thrown-away chromosomes, those remaining are chosen two-by-two
as parents to generate chromosomes of the child. There are different methods
for the parents’ selection. The tournament selection has been used in this article.
First, a small collection of chromosomes is accidentally selected in this method and
then the two chromosomes with the highest value are selected as parents.
(4) Genetic crossover: the parent chromosomes produce child chromosomes by
crossing over their genes. There are different methods of gene crossovers. Two
point crossovers are used in this article. Two parts of parents’ chromosomes are
selected in this method which can be seen in Figure 1. To produce a child
chromosome this way, gene-strings are copied from the beginning to the first
determined place from one parent, the gene-strings between two determined
places from the second parents, and the rest of the genes are copied from the
first parent.

Chromosome parent 1 Chromosome parent 2 Child Chromosome

Figure 1.
Two point crossover + =
 (5) Mutation: some genes of the new generation chromosomes (except the elite Fuzzy genetic
chromosome) are selected randomly and their value will be changed. This way, algorithm and
completely new chromosomes will be obtained.
(6) A return to the second step, and the repetition of steps until the stop condition is
neural networks
provided, which includes algorithm convergence and lack of change of the best
chromosome for the number of pre-defined generations, and the completion of
the pre-defined generation. The structure of genetics algorithm is as Figure 2. 285

Extracting rule from neural network

The major and main disadvantage of artificial neural networks is their disability in
explanation and analysis (Anderson, 1996). Neural networks are like continuous black

Criterion Function, Type

Chromosome, Genetic
Downloaded by UFRB At 18:36 18 October 2016 (PT)

Algorithm Parameters
and Operators

First Generation

Determining Value of Each

Chromosome

Parental Choice

Genetic Exchange

Mutation

Stop Conditions

Figure 2.
End Genetic algorithm
structure
 IJQRM boxes which make it difficult for neural network to perceive a solution (Mantas et al.,
31,3 2006). Therefore, the information used by neural network to achieve the solution is not
clear to the users and that may cause trouble in some cases (Huang and Xing, 2002).
To solve this problem, researchers tended to create a palpable and understandable
technique for neural network. They believe that they could achieve their goal by
extracting the produced rule of neural network (Huang and Xing, 2002). The purported
286 form of extracting a rule from the neural network is in the way (Figure 3).
We use genetic fuzzy system in this research to extract rule from neural network
whose basic concepts are defined in the following way:
.
Membership function. Every single input variable is transformed from the
numerical form to fuzzy form using fuzzy membership functions.
.
Information base. This section includes necessary information about input and
output variables and their governing rules. This section is itself constituted of
databases and rule bases.
.
Database. This section provides necessary definitions of membership functions
Downloaded by UFRB At 18:36 18 October 2016 (PT)

related to verbal terms and functions.

.
Rule bases. This section is constructed of rules in conditional sentence form of
“if-then” which are used to determine the output.
.
Inference system. In this section, the controlling output is determined regarding
fuzzy making input, the information of rule bases and using fuzzy concluding
methods.
.
Difuzzifier. Difuzzifier displays the output fuzzy set of conclusion system at a
non-fuzzy point. Actually, this section determines a point which is the best
representative of fuzzy collection.

The purported form of genetic fuzzy system is in the way (Figure 4).

Research methodology
This research is practical concerning its purpose and its analytical-descriptive
concerning its nature. Library methods such as books and essays have been used to
collect materials related to research background and also reports of the Stock Exchange
Corporation. Web sites have been used to gather data for research information.

Research statistical population

The daily price index of the Tehran Stock Exchange from 2005 to 2011 has been
selected as the statistical population; 1,383 data were accumulated for each variable
from related databases in the aforementioned period. The above-mentioned data were

Rule
Neural Extraction Rule Set
Network Algorithm
Figure 3.
Rule extraction structure
 Fuzzy genetic
algorithm and
Learning Process based on neural networks
Genetic Algorithm

Information Base 287

Data Base
Rule Base
(NN)

Inference
input Fuzzifier Difuzzifier output Figure 4.
System
Genetic algorithm system
Downloaded by UFRB At 18:36 18 October 2016 (PT)

divided into two groups, training and experimental, regarding the structure of neural
network. Each group consisted of 1,341 and 42 data, respectively, and it should also be
noted that mainly the accomplished prediction for the experimental period was selected
as the comparison criterion for models used in prediction.

Research goals
.
Designing a pattern to predict the price index of the Tehran Stock Exchange,
using fuzzy genetics algorithm.
.
Comparing conventional and intelligent methods in price index prediction and
determining an appropriate pattern for prediction.

Methods of price index modeling

In this study, we extracted 15 economic variables using (Gan et al., 2006; Ibrahim and
Aziz, 2003; Adam and Tweneboah, 2008; Gay, 2008) that explain the relationship
between economic variables and the stock market, using the co-integration Johnson
method considering the economic situation of Iran.
We performed principal component analysis (PCA) for refining data and elicited
four variables (S&P 500, gold price, oil price, exchange rate) that explain 89.3 percent of
changing these data. After that the stock market forecasting has been defined.
Methods of price index modeling will be explained in this section.

Neural network method

Preparing data is one of the complicated steps of neural network application, since the
best condition for neural network is when all the inputs and outputs are between 0 and 1.
One of the arguments for inserting inputs within range of 0 and 1 is the fact that transfer
functions (such as sigmoid function) are unable to differentiate between large amounts.
Therefore, the whole data were normalized using this formula: Xn ¼ X/Xmax.
Then, variables of S&P 500 index of the New York Stock Exchange, world gold price,
Iran basket crude oil price and state dollar value were determined as neural network
concerning studies of effective elements on price index. Prediction using MLP is in such
 IJQRM a way that the best output weight with the least prediction error is selected by imposing
31,3 training and learning on the network. After normalizing data, they were given to the
network. On the other hand, data were delivered to two groups in order to examine the
consistency of the output weight in such a way that first acquisition is done on training
data of prediction in order to examine the accuracy of network prediction. About
97 percent of the total used data were considered as training data and the rest were used
288 to examine the network. The amount of network learning by purpose functions was
continuously examined during the learning process and finally a network with the
fewest errors was selected. The parameters of the final neural network were determined
as shown in Table I.
In Figure 5, the prediction done by neural network has been compared to real values
for experimental data.
In addition, the values of examination criterion of functionality used for this model
is shown in Table II.
Concerning the amount of R 2 and adjusted R 2 obtained from the fitting model,
it can be concluded that the presented neural networks model is an appropriate one for
Downloaded by UFRB At 18:36 18 October 2016 (PT)

price index prediction. In such a way that the model with R 2 has been able to correctly
predict the trend of price index of the Tehran Stock Exchange for 95 percent.

Rule extracting method

Since fuzzy principles and rules have been used in this research to extract a rule with
the help of genetic algorithm and data exits and entrances, the first step is to introduce
membership functions to each variable. Triangular membership functions have been
used in this article. The triangular membership function which is used here consists of
five sizes of small (S), medium small (MS), medium (M), medium large (ML) and
large (L) which are shown in Figure 6.

Activation function Learning rate Hidden Epoch Mom

Table I.
MLP character Non-linear sigmoid 0/1 3 90,000 0/01

1.05
1
0.95
0.9
0.85
Figure 5. 0.8
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41
Neural network prediction
real NN

Table II. MSE MAPE R2 Adj. R 2

Examination criterion
of MLP 0/0048 0/0055 0/9586 0/9541
 Next step is to acquire a genetic algorithm to extract a set of rules. However, its Fuzzy genetic
necessary condition is to introduce the fitness function in order to begin learning the algorithm and
genetic algorithm and achieve appropriate rules. Fitness function is the least space
between the estimated output by genetic algorithm and MLP network output in this neural networks
research which is selected as the appropriate rules.
There are different parameters in genetic algorithm as well, in order to train and
learn genetic algorithm whose sizes are selected through trial and error methods to 289
reach appropriate rules. After learning with so much of these parameters, fuzzy genetic
algorithm was finally selected as the final model with the parameters (Table III).
The rules of examination criterion of functionality which are used for this model are
shown in below:
If X1 is ML and X2 is MS and X3 is M and X4 is M then Y is L.
If X1 is ML and X2 is MS and X3 is L and X4 is L then Y is L.
If X1 is L and X2 is ML and X3 is M and X4 is M then Y is ML.
Downloaded by UFRB At 18:36 18 October 2016 (PT)

If X1 is L and X2 is L and X3 is ML and X4 is ML then Y is ML.

If X1 is M and X2 is MS and X3 is ML and X4 is ML then Y is M.
If X1 is ML and X2 is M and X3 is ML and X4 is ML then Y is M.

S – small, MS – medium small, M – medium, ML – medium large, L – large.

After extracting the set of rules, we attempted to predict the price index relying on
experimental data in order to examine the function and accuracy of the prediction
techniques, so that it has predicted the trend of price index of the Tehran Stock
Exchange with an appropriate accuracy as is shown in Figure 7.
In addition, the values of examination criterion of functionality, which is used for
this model, is displayed in Table IV.
Regarding the value of R 2 and the adjusted R 2 obtained from fitting the model,
it can be concluded that the presented fuzzy genetic model is an appropriated model for
price index prediction, so that this model with R 2 has been able to correctly predict the
trend of price index of the Tehran Stock Exchange for 96 percent.

MS M ML
1

0.5 S L

Figure 6.
Membership size function
0 0.3 0.5 0.7 1

Chromosome size Iteration Swarm size Cross prob. Mutation prob. Membership size Table III.
Fuzzy genetic algorithm
150 350 5 0/5 0/7 5 parameter
 IJQRM Conclusion
31,3 The general goal of the research is to offer an appropriate model for price index
prediction of the Tehran Stock Exchange. Concerning the results of previous research,
which confirms the relative superiority of non-linear models in price index prediction, an
appropriate model has been offered in this research by compounding the non-linear
methods including nerve net and genetics algorithm with fuzzy theory in order to predict
290 the price index of the Tehran Stock Exchange. The fuzzy genetics algorithm used for
price index prediction of the Tehran Stock Exchange has had a better result, as shown in
Figure 8.
Regarding the R 2 examination criterion, these techniques also indicate its relative
superiority when compared to neural network technique. The extracted rules can help
the investors in their investment-related decision-making.

Suggestions for further researches

The following topics are suggested for further research regarding the significance of
prediction for managers and investors:
Downloaded by UFRB At 18:36 18 October 2016 (PT)

.
Predicting the rest of important financial factors such as earning per share (EPS),
dividend per share (DPS) and P/E ratio using neural network.
.
Comparing this method with fuzzy regression methods in price index prediction.
.
Using other types of neural networks in databases instead of fuzzy genetic
algorithm such as bee colony algorithm.

1.050
1.000
0.950
0.900
0.850
Figure 7. 0.800
Fuzzy genetic system 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41
prediction
REAL FORCAST GA

Table IV. MSE MAPE R2 Adj. R 2

Examination criterion of
fuzzy genetic algorithm 0/0042 0/0050 0/9673 0/9656

1.050
1.000
0.950
0.900
0.850
0.800
Figure 8. 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41
Comparison method used
REAL FORCAST GA FORCAST NN
 References Fuzzy genetic
Adam, A.M. and Tweneboah, G. (2008), “Macroeconomic factors and stock market movement: algorithm and
evidence from Ghana”, paper presented at Munich Personal RePEc Archive.
neural networks
Anderson, J.A. (1996), “Neural models with cognitive implications”, Basic Processes in Reading
Perception and Comprehension Models, Erlbaum, Hillsdale, NJ, pp. 27-90.
Chang, P.C. and Liu, C.H. (2008), “A TSK type fuzzy rule based system for stock price prediction”,
Expert Systems with Applications, Vol. 34, pp. 135-144.
291
Chen, T.L., Cheng, C.H. and Jong Teoh, H. (2007), “Fuzzy time-series based on Fibonacci sequence
for stock price forecasting”, Physica A, Vol. 380, pp. 377-390.
Chen, T.L., Cheng, C.H. and Jong Teoh, H. (2008), “High-order fuzzy timeseries based on
multi-period adaptation model for forecasting stock markets”, Physica A, Vol. 387,
pp. 876-888.
Cheng, C.H., Chen, T.L., Jong Teoh, H. and Chiang, C.H. (2008), “Fuzzy time-series based on
adaptive expectation model for TAIEX forecasting”, Expert Systems with Applications,
Vol. 34, pp. 1126-1132.
Downloaded by UFRB At 18:36 18 October 2016 (PT)

Chu, H.H., Cheng, C.H., Chen, T.L. and Haung, C.C. (2009), “Fuzzy dual-factor timeseries for stock
index forecasting”, Expert Systems with Applications, Vol. 39 No. 1, pp. 165-171.
Farnsworth, G.V., Kelly, J.A., Othling, A.S. and Pryor, R.J. (2004), Successful Technical Trading
Agents Using Genetic Programming, Sandia National Laboratories, Albuquerque, NM.
Gan, C., Lee, M., Yong, H.H.A. and Zhang, I. (2006), “Macroeconomic variables and stock market
interactions: New Zealand evidence”, Investment Management and Financial Innovations,
Vol. 3.
Gay, R.D. (2008), “Effect of macroeconomic variable on stock market returns for four emerging
economies: Brazil, Russia, India and China”, International Business and Economics
Research Journal, Vol. 7, p. 8.
Goldberg, D. (1989), Genetic Algorithms in Search, Optimization and Machine Learning,
Addison-Wesley, Reading, MA.
Hadavandi, E., Shavandi, H. and Ghanbari, A. (2010), “Integration of genetic fuzzy systems and
artificial neural networks for stock price forecasting”, Knowledge-Based System, Vol. 23,
pp. 800-808.
Haupt, R.L. and Haupt, S.E. (1980), Practical Genetic Algorithms, 2nd ed., Wiley, New York, NY.
Huang, S.H. and Xing, H. (2002), “Extract intelligible and concise fuzzy rules from neural
networks”, Fuzzy Sets and Systems, Vol. 132, pp. 233-243.
Ibrahim, M.H. and Aziz, H. (2003), “Macroeconomic variable and the Malaysian equity market:
a view through rolling subsamples”, Journal of Economic Studies, Vol. 30, p. 22.
Lin, L., Cao, L., Wang, J. and Zhang, Ch. (2007), The Application of Genetic Algorithms in Stock
Market Data Mining Optimization, Faculty of Information Technology, University of
Technology, Sydney.
Mantas, C.J., Puche, J.M. and Mantas, J.M. (2006), “Extraction of similarity based fuzzy rules
from artificial neural networks”, International Journal of Approximate Reasoning, Vol. 43,
pp. 202-221.
Refenes, A.N., Zapranis, A. and Francis, G. (1994), “Stock performance modeling using neural
networks: a comparative study with regression models”, Neural Network, Vol. 7 No. 2,
pp. 375-388.
 IJQRM Salcedo-Sanz, S., Fernández-Villacañas, J.L., Segovia-Vargas, M.J. and Bousoño-Calzón, C. (2005),
“Genetic programming for the prediction of insolvency in non-life insurance companies”,
31,3 Computer and Operation, Vol. 32 No. 4, pp. 749-765.
Tsang, P.M., Kwok, P., Choy, S.O., Kwan, R., Ng, S.C., Mak, J., Tsang, J., Koong, K. and
Wong, T.L. (2007), “Design and implementation of NN5 for Hong Kong stock price
forecasting”, Engineering Applications of Artificial Intelligence, Vol. 20, pp. 453-461.
292 Yoon, Y., Swales, G. and Margavio, T.M. (1993), “A comparison of discriminate analysis versus
artificial neural networks”, Journal of the Operations Research Society, Vol. 44 No. 1,
pp. 51-60.
Zadeh, L.A. (1994), “The role of fuzzy logic in modeling, identification and control”, Modeling
Identification and Control, pp. 191-203.

Further reading
Hsieh, T.J., Hsiao, H.F. and Yeh, W.C. (2010), “Forecasting stock market using wavelet transform
and neural networks: an integrated system based on artificial bee colony algorithm”,
Applied Soft Computing, Vol. 11 No. 2, pp. 2510-2525.
Downloaded by UFRB At 18:36 18 October 2016 (PT)

Karlaftis, M.G. and Vlahogianni, E.I. (2009), “Memory properties and fractional integration in
transportation time-series”, Transportation Research Part C, Vol. 17, pp. 444-453.
Kasman, A., Kasman, S. and Torun, E. (2009), “Dual long memory property in returns and
volatility: evidence from the CEE countries’ stock markets”, Emerging Markets Review,
Vol. 10, pp. 122-139.

Corresponding author
Noraddin Mousazadeh Abbassi can be contacted at: noraddinm@gmail.com

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com

Or visit our web site for further details: www.emeraldinsight.com/reprints