Australian Journal of Basic and Applied Sciences

Transcription

1 AENSI Jourals Australia Joural of Basic ad Applied Scieces ISSN: Joural home page: Stock Price Predictio of Oil ad Gas Corporatio usig Modified Geetic Algorithm Simulated Aealig Approach 1 S.Kopperudevi ad 2 DR.A.Iyemperumal 1 Research Scholar, Dr.M.G.R.Educatioal Ad Research Istitute,Uiversity, Cheai, Tamil Nadu, Idia. 2 Professor, Departmet of Mathematics, Dr.M.G.R.Educatioal Ad Research Istitute,Ursity, Cheai, Tamil Nadu, Idia. A R T I C L E I N F O Article history: Received 2 March 2014 Received i revised form 13 May 2014 Accepted 28 May 2014 Available olie 23 Jue 2014 Keywords: Artificial eural etwork, Geetic Algorithm, Simulated Aealig, Stock Price A B S T R A C T Backgroud:Stock Market is amessy place for predictig sice there are o sigificat rules to estimate or predict the price of shares i the stock market. May methods like techical aalysis, fudametal aalysis, time series aalysis ad statistical aalysis, etc. are all used to attempt to forecast the price i the stock market, but oe of these methods are prove as a cosistetly acceptable predictio tool. Objective:I this paper, a artificial eural etwork based o Modified Geetic Algorithm-Simulated Aealig (MGASA) is used to predict the stock price idex.i desigig the model, the data of oil ad gas compay is take from Bombay Stock Exchage (BSE) ( ).Result: The etwork is traied by 60% of the experimetal data. 30% of the essetial iformatio which had bee ackowledged for testig the appropriateess has bee fed ito the model. The predicted values were compared with the experimetal values for evaluatig the performace. The result obtaied by usig MGASA are i astoudig cocurrece with the experimetal results ad has high executio i stock price predictio.coclusio:it is observed that the proposed algorithm sigificatly outperforms resultig i more profits. Hece, it ca be cocluded that the proposed algorithm is well suited for predictio of the stock prices AENSI Publisher All rights reserved. To Cite This Article: S.Kopperudevi ad DR.A.Iyemperumal., Stock Price Predictio of Oil ad Gas Corporatio usig Modified Geetic Algorithm Simulated Aealig Approach. Aust. J. Basic & Appl. Sci., 8(9): , 2014 INTRODUCTION Stock market statistics play a vital role i hypothetical research, especially i the past decades. A importat hypothesis related to the stock market, which has bee debated ad researched time ad agai is EMH (Efficiet Market Hypothesis). Accordig to the EMH, the stock market immediately reflects all of the iformatio available publicly. But i reality, the stock market is ot that efficiet, so the predictio of stock market is possible. Stock market predictio is a act of attemptig to determie the upcomig value of a busiess stock traded o fiacial exchage. This would yield a cosiderable profit i busiess. May methods like techical aalysis, fudametal aalysis, ad time series aalysis are used to predict stock market price, where time series aalysis seeks to determie the future price of a stock based oly o the probable past price. Researchers aalyzed ad faced that the stadard time series models have umerous drawbacks i precisio ad robustess. There is o experimetal evidece of liearity i stock returs, various researchers ad fiacial experts have focused o the oliear predictio methods. The hybrid GS etwork as a potetial stock market aalysis tool, which is a combiatio of Neural Network ad Geetic Algorithm, has bee used by Robert Verer for stock market predictio, a field of Artificial Itelligece is capable of providig a better result i predictig the fiacial market to help fiace practitioers to make qualitative decisios but it has a overhead of higher computatioal requiremets ad time. This paper proposes Modified Geetic Algorithm ad Simulated Aealig Network (MGSN) ad applied to Oil ad Gas Stock Price Predictio ad produces higher quality solutios ad overheads less computatio time. The results obtaied from these applicatios have proved that the MGSN has the ability of addressig large ad complex problems ad miimizig the MSE error value i traiig ad testig period by improvig a accuracy i Oil ad Gas Stock Price Predictio ad is a ew promisig predictio algorithm for stock market forecastig. Related Works: Correspodig Author: S.Kopperudevi, Research Scholar, Dr.M.G.R.Educatioal Ad Research Istitute,Uiversity, Cheai, Tamil Nadu, Idia.

2 376 S.Kopperudevi ad DR.A.Iyemperumal, 2014 Computers play a vital role i each ad every field, especially i stock markets. Before the ivetio of computers, Shareholders ad Fiaciers iitially forecasted stocks based o their ituitio. This helped fiacial practitioers to make decisios o price predictio o stock values. The vast improvemet i tradig stocks ad shares force us to fid a better mechaism, with the help of computers, to predict the stock price i a short period of time with more accuracy i order to uptur profits, thus dimiish losses. Several research efforts were carried out to observe the forecasted price i stock markets. Various techiques like fudametal aalysis, techical aalysis, ad time series aalysis, data miig techiques, machie learig algorithms, chaos theory ad liear regressio ad machie learig algorithms have bee used to predict stock market data. Researchers aalyzed ad faced that the above metioed models have umerous shortcomigs i precisio ad robustess of statistics. The techical aalysis ad fudametal aalysis take a log time to respod to a compay about stock price. Time series aalysis seeks to determie the future price of a stock based oly o the probable past price. There is o experimetal evidece of liearity i stock returs, various researchers ad fiacial experts have focused o the oliear predictio methods. Hybridized approach, a data miig techique, improved approach of techical ad fudametal aalysis provides ehaced accuracy of stock predictio, which is ot attempted to fix the critical effect of specific aalysis variables. Support Vector Machie (SVM) (Huag et al., 2005) ad Reiforcemet Learig, a Machie Learig Algorithms (Vatsal H. Shah), which are iteded to accumulate data from umerous global fiacial markets makes the algorithm slower to calculate the immiet price of stock. Time Delay, Recurret, ad Probabilistic Neural Networks have certai disadvatages like executio complexity, shortage of memory ad require much time for testig where each method is used to predict forthcomig value of a stock based o the history of dayto-day closig prices. Artificial Neural Networks (ANNs) ad Geetic Algorithms, a field of Artificial Itelligece is capable of providig a better result i predictig the fiacial market to help fiace practitioers to make qualitative decisio. There is a wide variety of research work o the applicatios of Neural Networks especially i fiace ad stock markets. Artificial eural etworks are competet iaccurate predictios without ay specific assumptios about variables ad their effectiveess. (Abdüsselam Altukayak,2009) utilized a geetic algorithm for the forecastig of sedimet load ad discharge. Very few have attempted to utilize just geetic algorithms to foresee stock prices. Sice the geetic algorithm ca perform sesibly well by ad large there must be a approach to aticipate stock price utilizig GA. Shaikh A. Hamid ad Zahid Iqbal preset a preparatio for utilizig eural etworks for fiacial determiig. They aalyze istability estimates from eural etworks with iferred upredictability from S&p 500 Idex future alteratives utilizig the Baroe-Adesi ad Whaley (BAW) America future alteratives estimatig model. Gauges from eural etworks beat itimated upredictability gauges ad are most certaily ot discovered to be essetially uique i relatio to ackowledged upredictability (Shaikh, 2003). (David Eke ad Surapha Thaworwog,2005) Presets a iformatio gai procedure utilized as a part of machie learig for iformatio miig to assess the presciet coectios of various fiace related ad ivestmet variables. Neural system models for level estimatio ad groupig are the aalyzed for their capability to give a compellig gauge of future qualities. (Zhag Yudog ad Wu Lea,2008) proposed a improved bacterial chemo taxis ehacemet (IBCO), which is the icorporated ito the back propagatio (BP) artificial eural system to create a efficiet aticipatig model for expectatio of differet stock records. Experimets demostrate to its better performace over other systems i takig i capacity ad geeralizatio. (E.L. de Faria ad J.L. Gozalez, 2009) performs a predictive ivestigatio of the chief idex of the Brazilia stock market through artificialeural etworks ad the versatile expoetial smoothig strategy. The target is to compare the aticipatig executio of both systems o this market record, also, specifically, to assess the exactess of both systems to predict the idicatio of the market returs. Additioally the impact o the outcomes of a few parameters associated with both systems is cotemplated. Their effects demostrate that both systems produce comparative outcomes i regards to the predictio of the record returs. O the opposite, the eural etworks outperform the adjustable expoetial smoothig strategy i the gaugig of the market developmet, with relative hit rates like the oes foud i other created markets. (E.L. de Faria ad J.L. Gozalez, 2009)Performs a presciet ivestigatio of the pricipal idex of the Brazilia stock exchage through artificial eural etworks ad the versatile expoetial smoothig strategy. The target is to compare the aticipatig executio of both routies o this busiess sector idex, also, specifically, to assess the correctess of both techiques to aticipate the idicatio of the busiess returs. Likewise the impact o the outcomes of a few parameters chatted to both techiques is cosidered. Their effects demostrate that both strategies produce comparative outcomes with respect to the predictio of the idex returs. O the opposite, the eural etworks beat the versatile expoetial smoothig strategy i the determiig of the busiess developmet, with relative hit rates like the oes foud i other developed markets.

3 377 S.Kopperudevi ad DR.A.Iyemperumal, 2014 Fiacial forecastig is of respectable pragmatic ivestmet furthermore, because of the artificial eural etwork's capability to mie profitable data from a mass history of iformatio; its provisios for fiacial estimatig have bee extremely promiet i the course of the last few years (T. H. Roh, 2007). (Gurese, et al., 2011)Reported the legitimacy of ANNs i stock busiess idex predictio. Sheg-Hsu Hsu ad JJ Po-A Hsieh study utilizes a two-stage desig for better stock price predictio. Particularly, the self-orgaizig map (SOM) is iitially used to deteriorate the etire iformatio space i areas where iformatio focuses with comparable factual circulatios are gathered together, i order to hold ad catch the o-statioary property of fiacial arragemet. I the wake of breakig dow heterogeeous iformatio focuses ito a few homogeous districts, support vector regressio (SVR) is coected to predict fiacial idices. The proposed system is experimetally tried utilizig stock price arragemet from seve sigificat fiacial markets(sheg-hsu Hsu ad JJ Po-A Hsieh, 2008). The mai objective of this paper ivolves i attemptig to predict the itrisic value of Oil ad Gas i Stock market. Followig techiques forecast performace differeces amog differet types of models ad eural etwork. We itroduced a ew model, combiatio of Geetic Algorithm (GA) ad Simulated Aealig (SA), Modified G-S etwork for Oil ad Gas Price Predictio, to improve o the existig approaches of forecastig the upcomig value of Oil ad Gas. Geetic Algorithm is a experimetal scrutiy which provides the best solutio i specific time. Simulated Aealig is a effective techique to obtai a cosiderable future Oil ad Gas stock price by a specified amout of pride. But it fails i providig optimal solutio. While combiig these two algorithms, we ca be able to fid a great solutio to predict stock price value with miimal time for a specific period irrespective of icreasig time period. This will improve the solutio preseted here. The stock price is chaged time to time i microsecods, where it is more importat to predict accurate values of future price to get to profit i the stock exchage. This G-S Network allows cotemplative aalysis of small ad large set of statistics, especially those that have the tedecy to oscillate withi a short of period of time. The performace of this method is compared with other techiques. The Modified G-S Network would be a best approach rather tha a time series aalysis, curret Neural Networks ad other methods. However, the focus of this paper will improve accuracy i Oil ad Gas Stock Price Predictio with a short period of time. Basics of Geetic Algorithm ad simulated aealig: Geetic Algorithm: Geetic algorithms (GA) are a particular kid of Evolutioary Algorithm (EA).The essetial priciples of Geetic Algorithms (GAs) were proposed by Hollad i 1975 (Hollad JH,1975). GAs are optimizatio ad search procedure that are based o the mechaics of biological evolutio. They have bee applied successfully to solve a variety of complex problems (Beasley, D, Bull, D R ad Marti, R, 2008). I geeral geetic algorithm works as follows: The geeral sketch of GA i pseudocode Algorithm: GA(, a,α): i:=0; p i =populatio of radomly selected idividuals; compute fitess(x) for each x p i ; //Iitialize geeratio //Iitialize geeratio //Evaluate P i do: 1.select: Select(1-a) members from p i ad isert ito p i+1 ; 2.crossover: Select a members from p i ; pair them upo ad produce offsprig; isert them ito ito p i+1 ; 3.mutatio: Select α members from p i+1, ivert a radomly selected bit 4.Evauate p i+1 ; Compute fitess(x); icremeti=i+1 while fitess(i) ot high eough; retur fittest idividual from p i The algorithm starts with geeratig iitial populatio radomly. Idividuals from the populatio are selected for reproductio based o their fitess value. The selected chromosomes are recombied (crossover) ad mutated to geerate ew populatio. The process is cotiued util a termiatio coditio is met.

4 378 S.Kopperudevi ad DR.A.Iyemperumal, 2014 Simulated Aealig: Simulated aealig (SA) is a radom-search techique [10] for combiatorial optimizatio problems to search for feasible solutio ad coverge to a optimal solutio. The idea of SA is based o thermodyamics, process of coolig metals (aealig). Whe you heat metal at a meltig poit ad the gradually cooled, a large crystals will be formed. If the fluid is cooled quickly the crystal will cotai blemishes (Kirkpatrick., 1983). The SA performs a radom search o the rage of values with metropolis criteria. The performace of SA is based o the aealig schedule. Simulated aealig is a straightforward algorithm for a set of optimizatio heuristic that searches for a optimal eighborhood solutio. The major beefit of SA over other traditioal local search techiques is that its potetial to escape from local miima. The basic priciple of SA is as follows: Geerate iitial solutio S p Set iitial temperature t 0 Set 0 < β < 1 Loop: Select eighborig solutio x i Evaluate f(x i ) Calculate δf = f(x i ) - f(x j ) if(δf< 0) Thex j = x i Else if 1 1+e δf t > radom(0,1) Thex j = x i Else t( k+1) = β t(k) util termiatio coditio is met. The algorithm starts with a iitial solutio. It the selects the eighborhood solutio ad evaluates the objective fuctio. The value of the objective fuctio is better tha the curret solutio, the it is accepted. It also accepts worse quality solutios based o some probability. The process cotiues util the termiatio coditio is met (Roh, T.H., 2007). Modified Geetic Algorithm ad Simulated Aealig: Geetic algorithms ca save brilliat idividuals for the followig geeratio i the geetic operatio process ad assure the assorted qualities of the populatio. The simulated aealig algorithm has the strog local search capability ad is equipped for gettig away from local optimal solutios. Ayway GAs is proe to premature covergece ad be trapped i local optimal solutios. Likewise, the SA requires more reckoig time. Thus, by the sythesis of the two algorithms, amodified Geetic Algorithm-Simulated Aealig algorithm is demostrated i this area. I geeral a allied methodology of GAs ad the SA is to house the SA iside GAs. The SA ehaces each idividual from GAs populatios with a iteratio umber that is obliged to achieve Markov chai legth. Alog these lies, the accepted GA-SA takes sigificatly more executio time tha GAs or the SA. To defeat this iadequacy, this study ehaces the customary GA-SA algorithm. The ehaced algorithm chages the optimal method of the SA to the GAs populatio, that is, the SA just ehaces the optimal idividual of GAs populatio, ot all people. After the chage, the algorithm ca spare substatially more executio time tha the customary GA-SA. Additioally, the MGASA is equipped for attaiig better results tha other improvemet strategies. MGASA Algorithm: GA Phase: Step 1: Iitialize populatio ad temperature. Step 2: Evaluate the populatio Step 3: Repeat Apply selectio operator Apply crossover operator Apply mutatio operator Evaluate populatio Util termiatio coditio is met

5 379 S.Kopperudevi ad DR.A.Iyemperumal, 2014 SA, Phase: Step 4: Select best optimal solutio from GA Step 5: Evaluate the objective fuctio. Step 6: Repeat Geerate ew eighbourhood solutio Estimate fitess fuctio Accept ew eighbourhood based o metropolis criteria Util ( max solutios to be cosidered for each sigle iteratio) Step 7: Decrease the temperature usig the aealig schedule. Step 8: Repeat steps 6-7 util stoppig criteria is met. The MGASA algorithm comprises of two stages, the GA stage ad the SA stage. I the IGA-SA algorithm, Iitially GAs creates the iitial populatio radomly. The GA the assesses the iitialpopulatio ad works o the populatio utilizig three geetic operators to process ew populatio. After every geeratio the GA seds the best idividual to the SA i phase II for further chage. Havig completed the further chage of the idividual, the SA seds it to the GA for the followig geeratio oce more. This methodology proceeds util the termiatio coditio of the algorithm is met. Phase 1 Optimal geetic algorithm process: The GA produces stochastically the iitial populatio ad afterward operates o the populatio utilizig three geetic operators to prepare ew populatio. As per pseudo code of the geetic algorithm, a few parts i respect to GAs ought to be resolved, for example, the choice variables, the populatio estimate, the geeratio ofthe iitial populatio, the assessmet of populatio, the plas of ecodig ad iterpretig for chromosomes, the determiatio of geetic operators ad the termiatio coditio. Objective Fuctio: The objective is to decrease the forecastig error of oil ad gas stock price. The objective fuctio ca be writte as: RMSE = i=1 A P 2 Where is the populatio size, A is the actual price ad P is the predicted value. Geerate Iitial Populatio: The iitial populatio is produced radomly. Each of Iitial weights is radomly created betwee -1 ad + 1. Fitess Fuctio: GAs assesses the populatio depedet upo the fitess fuctio. A idividual with higher fitess rate has higher opportuity to be chose ito the followig geeratio. Geerally the fitess of a strig is with respect to the target fuctio. A P 2 RMSE = Selectio Procedure: We utilize trucatio selectio for selectig the populatio. I trucatio selectio people are sorted as per their fitess. Just the best idividuals are chose for idividuals. The trucatio limit shows the extet of the populatio to be chose as idividuals. At that poit we utilize a biary trucatio selectiofor producig ew offsprigs by utilizatio of geetic operators. I trucatio selectio, two members of the populatio are chose as arbitrary ad their fitess cotrasted ad the best oe cocurrig with fitess worth will be decided to oe paret. Likewise a alterate paret chose with the same techique. Geetic Operators: Here, we utilize two-poit crossover ad oe-poit mutatio as geetic operators. Replacemet: The preset populatio has bee replaced by the recetly produced offsprigs, which structures the ext geeratio. Termiatio Criteria: If the umber of geeratio equivalets the maximum geeratio umber the stop. i=1

6 380 S.Kopperudevi ad DR.A.Iyemperumal, 2014 Phase 2 Optimal Simulated Aealig Process: I the methodology of the MGASA, the GA will sed its best idividual to the SA for ehacemet. After the optimal idividual of the GA beig ehaced, the SA passes it to the GA for the subsequet geeratio. This methodology proceeds util the termiatio coditio is met. Iitial Temperature: The SA accepts ew states depedet upo Metropolis criterio which is a stochastic procedure. The criterio is give by P(e)= mi{1, exp( δe/t)}, where δe =f(s i ) f(s j ) is the differece of the objective fuctio values of the ew state s i ad the preset state s j, ad t is the preset temperature. Assumig that δe is ot exactly zero, the the ew state is held ad the preset state is discarded. Overall, the ew state may be held if the Boltzma likelihood,p b =exp( δe/t), is greater tha a arbitrary umber withi the rage 0 to 1. At a high temperature, the SA ca accept aother state that has a higher value tha that of the past uified with a substatial likelihood. As coolig proceeds, the state may be accepted by the SA with a less likelihood. Coolig Rate: The performace of the SA is relative with respect to the coolig rate. So as to ehace the cosistecy ad the search effectiveess of the SA, a eormous coolig rate ought to be maitaied. I the evet that the coolig rate of each temperature chage couter is excessively low, the SA will cost reckoig time expediture. O alterate hads, if a faster coolig rate is utilized, the likelihood of gettig trapped ito a local miimum is higher. I geeral, the value of coolig rate may be cotrolled by its sesitivity aalysis. The coolig schedule is give as follows: T k = γ T k-1 Where T k ad T k-1 are temperatures at time k ad k-1; γ is the coolig rate betwee 0 ad 1. Number of Trasitios at a Temperature: I a search methodology of the SA, the state move at every temperature chage couter is just depedet o the ew states ad curret status. Hece, the search procedure of SA could be ackowledged as a Markov chai, whose legth is characterized by the amout of moves permitted at the curret temperature. The amout of moves at each temperature is characterized as: R t = αt R is the maximum umber of repetitios at a particular temperature, α is a costat variable. Geeratio of eighbourhood structure: The focus of eighbourhood structure geeratio is to chage arbitrarily the preset state to a feasible rage of its curret value. There are umerous diverse approaches to geerate the eighborhood structure. I the preset work, the o-uiform trasformatio approach i the GA is received with some adjustmet for geeratio methodology. I the evet that a uiform arbitrary umber distributed i the rage [0,1] is less tha the mutatio Pm, the preset choice variable is permitted to trasform its value radomly. Otherwise, the preset decisio variable is ot permitted to do that. Termiatio Coditio: The algorithm rus util the last geeratio or whe the low RMSE value is reached. Simulatio Study: Experimetal Data: The research iformatio utilized withi this study is BSE oil ad gas stock idex from 1 Jauary 2010 to 31 December We gather a sample of 48 tradig moths; we pick 60% for traiig phase ad 30% for the testig phase. Numerous past stock market ivestigatios have utilized techical idicators as characteristics. Techical idicators are compoets that forecast the future performace of stocks i a give set of ecoomic situatios. By ad large techical idicators are utilized for short term desigs. They are regularly depedet upo scietific estimatios which take ito cosideratio the curret relatioship betwee the stock price ad the geeral developmet of the market where the stock is exchaged. These idicators are ascertaied depedet upo fudametal qualities: closig price, opeig price, high price, low price, all these prices speak to the stock quality throughout the tradig sessio. I this research, we utilize the techical idicators as iput variables. We pick seve techical idicators to costrictiothe set of variables. These are calculated from the raw data as demostrated (RitajaliMajhi et al., 2008).

7 381 S.Kopperudevi ad DR.A.Iyemperumal, 2014 Performace Evaluatio: Traiig of the forecastig models utilize MGASA algorithm.the, utilizig these weights the same aticipatig models are agai utilized for the testig reaso. The assessmet is doe to test the executio of the model for forecastig the close price of the idex. The Mea Squared Error(MSE), Root Mea Squared Error(RMSE), R-Squared(R 2 ), Adjusted R- squared(r A 2 ), Haa-Qui Iformatio Criterio (HQ) are used to gauge the performace of the traied forecastig model for the test data(table 1). Table 1: Performace Criteria ad the related formula. Performace Criteria Mea Squared Error Root Mea Squared Error (RMSE) R-Squared(R 2 ) Adjusted R-Squared(R 2 A ) Haa-Qui Iformatio Criterio (HQ) Formula MSE = y 1 y 2 2 RMSE = i=1 i=1 i=1 y 1 y 2 2 y R 2 1 y 2 2 = i=1 y y 2 2 y = real value, y 1 = estimated value, y 2 = mea value 2 R A = 1 (1 R 2 ) T 1 T HQ = l SSR k l [l ] + SSR = i=1 y y 2 1 Results: I this paper data from to are utilized for traiig purpose ad the predict the stock close price of the year 2013 i.e, from Jauary 2013 to December 201 ad compare it with the closig data of that year. Table 2: Actual ad predicted price usig modified geetic algorithm simulated aealig. Period(2013) Actual Predicted Jauary February March April May Jue July August September October November December Actual Predicted 75 Ja Mar May July Sep Nov Fig. 1: Actual ad predicted price usig modified geetic algorithm simulated aealig. Table 2 represets the actual value ad the predicted value of the proposed approach. Fig 1 represets the test results by plottig the actual value agaist the value predicted by usig the proposed algorithm.

8 382 S.Kopperudevi ad DR.A.Iyemperumal, 2014 Table 3: Error rate of the proposed algorithm usig various test criteria. Test Criteria Error Rate(%) Mea Squared Error 3.45 Root Mea Squared Error (RMSE) 5.48 R-Squared(R 2 ) 0.17 Adjusted R-Squared(R 2 A ) 1.15 Haa-Qui Iformatio Criterio (HQ) Table 3 Shows the error rate of the proposed techique by usig various methods. The proposed algorithm performed the predictio better tha the other ivestigated model. Coclusio: Now-a-days Oil ad Gas corporatio because of the icrease i ifrastructure ivestmet, is a attractive market for ivestmet. Thus modellig a framework for stock price predictio i oil ad gas corporatio is vital for all traders ad fiacial cosultats to decrease their risk ad icrease the beefit of the shareholders. I this research, a Modified Geetic Algorithm-Simulated Aealig is used to predict the stock price of Oil ad Gas Corporatio take from Bombay Stock Exchage. The stock prices are estimated by the proposed MGASA algorithm ad the effectiveess of the proposed algorithm was validated o the origial data. It is observed that the proposed algorithm sigificatly outperforms resultig i more profits. Hece, it ca be cocluded that the proposed algorithm is well suited for predictio of the stock prices. REFERENCES Abdüsselam Altukayak, Sedimet load predictio by geetic algorithms Advaces i Egieerig Software, 40(9): Beasley, D., D.R. Bull, ad R. Marti, A Overview of Geetic Algorithms:. Part 1, Fudametals. Norwegia Uiversity of Sciece ad Techology. David Eke ad Surapha Thaworwog, The use of data miig ad eural etworks for forecastig stock market returs, Joural of Fiace, USA. De,. E.L., Faria ad J.L. Gozalez, Predictig the Brazilia stock market through eural etworks ad adaptive expoetial smoothig methods, Expert Systems with Applicatios Article i Press. Gurese, E., G. Kayakutlu, T.U. Daim, Usig artificial eural etwork models i stock market idex predictio, Expert Systems with Applicatios, 38(8): Hollad, J.H., Adaptatio i atural ad artificial system. The Uiversity of Michiga Press, A Arbor, MI. Huag, W., et al., Forecastig stock market movemet directio with support vector machie, Computers & Operatios Research, 32: Kirkpatrick, S., C.D. Gelatt, M.P. Vecchi, Optimizatio by simulated aealig. Sciece, 220: RitajaliMajhi, G., Pada, G. Sahoo, Abhishek Pada, ArvidChoubey, predictio of S&P500 ad DJIA Stock Idices usig Particle Swarm Optimizatio Techique IEEE. Roh, T.H., Forecastig the Volatility of Stock Price Idex, Joural of Expert Systems with Applicatios, 33: Shaikh, A., Hamid ad Zahid Iqbal, Usig eural etworks for forecastig volatility of S&P 500 Idex futures prices, School of Busiess, USA. Zhag Yudog ad Wu Lea, Stock market predictio of S&P 500 via combiatio of improved BCO approach ad BP eural etwork, Expert Systems with ApplicatiosVolume 36, Issue 5, Pages Sheg-Hsu Hsu ad J.J. Po-A Hsieh, A two-stage architecture for stock price forecastig by itegratig selforgaizig orgaizig map ad support vector regressio, Expert Systems with Applicatios, 36 (4): Vatsal, H., Shah Machie learig techiques for stock predictio,