prediction of stock using artificial neural networks (a survey)

International Conference on Computing and Intelligence Systems Volume: 04, Special Issue: March 2015 Pages: 1208 – 1212 ISSN: 2278-2397

International Journal of Computing Algorithm (IJCOA) 1208

Prediction of Stock using Artificial Neural Networks (A Survey)

G.Sundar1, K.Satyanarayana2

1Research Scholar, Bharath University, Chennai Principal &Prof, Dept of Computer Science,Sindhi College,Chennai

Email: [email protected], [email protected]

Abstract-Supreme prediction of stock price revolutionizes is vastly tricky and momentous information for investors. The use of artificial neural networks has found a variegated field of applications in the present world. This has led to the development of various models for financial markets and investment. To outline the design of Artificial Neural Network model with its features and various parameters. Investors need to understand that stock price data is the most important information which is highly impulsive,a non-parametric and are affected by many suspicions an interrelated economic and political factor across the globe. Neural Networks can be used as a better substitute technique for forecasting the daily stock market prices. This study demonstrates the literature analysis of prophecy in stock market.

Keywords: Artificial Neural Networks (ANNs), Forecasting, Stock Price, Prediction methods

I. INTRODUCTION

Artificial Neural Network(ANN)conceptions used over the last few years and applied to different problem domains as finance, medicine,manufacturing,geology and physics. Stock market is one of the major investments for backer for the past 20 years. The Bombay Stock Exchange (BSE) is one of the India’s major stock exchange and the oldest exchanges across the world. To guess the closing price of stock market with reference to some companies listed under BSE using artificial neural network many methods have been made on stock market to predict the stock prices. ANN helped researcher in developing many model using deep-rooted, technical and time series in forecasting the competitive predictions of stock price for the investors.The ANNs have the ability to determine the nonlinear relationship in the input data set without a priori assumption of familiarity of relation between the input and theoutput. Hence ANNs match well than any other reproduction in predictthe stock price.

Neural Network is extremely appropriate for resembling Fuzzy Controllers and other types of Fuzzy Expert Systems. The concept of Artificial Neural Network (ANN) begins with the functions of human brain. Human brain consists of approximately 10 billion neurons (nerve cells) and 6000 times as

many connections between them. ANN as motivated by biological nervous systems, process information supplied, with a large number of highly interconnected processing neurons. ANN may be considered as a data processing technique that maps or relates some type of input stream of information to an output stream of data.

ANN consists of neurons (nodes) distributed across layers. The way these neurons are distributed and connected with each other defines the architecture of the network. The connection between the neurons is attributed by a weight. Each neuron is a processing unit that takes a number of inputs and after processing with an activation function called transfer function yields a distinct output. The most commonly used transfer functions include, the hardlimit, the purelinear, the sigmoid and the tan sigmoid function.ANNs are considered nonlinear statistical data modeling tools where the complex relationships between inputs and outputs are modeled or patterns are found.

II. ARTIFICIAL NEURON NETWORK

An artificial neuron network (ANN) is a computational model based on the structure and functions of biological neural networks. Information that flows through the network affects the structure of the ANN because a neural network changes - or learns, in a sense - based on that input and output.

Figure 1 Formation of ANN

In Figure 1 various inputs to the network are represented by the mathematical symbol, x(n). Each of



these inputs is multiplied by a connection weight. These weights are represented by w(n). In the simplest case, these products are simply summed, fed through a transfer function to generate a result, and then output. This process lends itself to physical implementation on a large scale in a small package. This electronic implementation is still possible with other network structures which utilize different summing functions as well as different transfer functions.

A. Layers in ANN

ANNs have three layers that are interconnected. The first layer consists of input neurons. Those neurons send data on to the second layer, which in turn sends the output neurons to the third layer.

Figure 2 Layers in ANN

Figure. 2A shows a neuron. The activation function is usually non linear, with a sigmoid shape (e.g., logistic or hyperbolic tangent function). A simple network having the universal approximation property (i.e., the capability of approximating a non linear map as precisely as needed, by increasing the number of parameters) is the feed forward MLP(Multi Layer Perceptron) with a single hidden layer, shown in Figure. 2B.

B. Methods of ANN Learning Learning based solutions can becategorized as:

Supervised or associative learning. Where the net is trained by quantifying input, as well as matching output patterns.These input/output pairs may either be provided by an external teaching component, or by the net itself (self-supervisedapproach). Unsupervised learning (self-organizing paradigm). Where the net (output) unit is trained to respond to clusters ofpattern within the input framework. In this paradigm, the system is supposed to discover statistically salient features ofthe input population. Compared to the supervised learning method, there is no a priori set of categories into which thepatterns are to be classified, rather the system has to develop its own representation of the input stimuli. For supervisedand unsupervised learning methods, basically all the learning rules reflect a variant of the Hebbian learning rule.

Reinforcement Learning. Where the net applies a learning paradigm that is considered an intermediate form of theabove 2 types of learning. In this method,

the learning machine executes some action on the environment, and as aresult, receives some feedback/response. The learning component grades its action (as either good or bad) based on theenvironmental response, and adjusts its parameters accordingly. Generally speaking, the parameter adjustment processis continued until an equilibrium state surfaces where no further adjustments are necessary

Back-propagation and the conjugate-gradient method(also known as MLP –Multi Layer Perceptron) work by iteratively training the network using the presented training data. On each iteration (or epoch), the entire training set is presented to the network, one case at a time. In order to update the weights a cost function is determined in terms of the weights and its derivative (or gradient) with respect to each weight is estimated [10]. Weights are updated following the direction of steepest descent of the cost function. A common cost function and the one used in this work is the Root Mean Squared ( RMS) error. During experimentation with the two methods back-propagation was a little slower to learn that the conjugate-gradient approach, but both methods resulted in almost identical error rates. Back-propagation was slightly more accurate, making one more correct prediction than did conjugate-gradient.

C.Basic Features of ANN 1. High computational rates due to massive parallel

processing. 2. Fault tolerance is high as damage to few nodes

doesn’t significantly effect the overall performance.

3. With Learning and Training feature the network adopts itself based on the information received by it in the past environment.

4. ANN’s feature Goal seeking where the performance to achieve the goal is measured and is used to self organizes the system.

5. They are Primitive Computational Elements as each element resembles one simple logical neuron and it cannot do much.

D. Objective of Neural Network:

The artificial neural network is composed of many artificial neurons that are linked together according to specific network architecture. The objective of the neural network is to transform the inputs into meaningful outputs.

Figure 3 Objective of Neural Network



The artificial neuron given in figure 3 has N input, denoted as x1, x2,…..xm. Each line connecting these inputs to the neuron is assigned a weight, which is denoted as w1, w2, ….wmrespectively. Weights in the artificial model correspond to the synaptic connections in biological neurons. The threshold in artificial neuron is usually represented by θ and the activation Фcorresponding to the graded potential is given by the formula:

III. STOCK MARKET PREDICTION

It is the act of trying to determine the future value of a company stock or other financial instrument traded on an exchange. The successful prediction of a stock's future price could yield significant profit. The efficient-market hypothesis suggests that stock price movements are governed by the random walk hypothesis and thus are inherently unpredictable. Others disagree and those with this viewpoint possess a myriad of methods and technologies which purportedly allow them to gain future price information. When applied to a particular financial instrument, the random walk hypothesis states that the price of this instrument is governed by a random walk and hence is unpredictable. If the random walk hypothesis is false then there will exist some (potentially non-linear) correlation between the instrument price and some other indicator(s) such as trading volumeor the previous day's instrument closing price. If this correlation can be determined then a potential profit can be made.

A stock market is a public market for trading the company’s stocks and derivative at an approved stock price. These are called securities, listed on a stock exchange as well as an investor traded privately. In the stock market also known as secondary market is monitored by a regulatory body called SEBI (Security and Exchange Board of India). It depends upon the demand and supplies the prices are vary. The price will high when the demand is high, when the share is heavy to sell the decrease the price. This type of transaction is called trading and the companies, which are permitted to do the trading, are called “Listed companies”.

IV. USING ANNS IN FORECASTING STOCK

PRICES

An Artificial Neural Network is foundationon a data processing system consisting of a large number of simple highly interconnected processing elements in an architecture inspired by the structure of the cerebral cortex of the brain. A neural network is a extraordinarilysimilarscattered process that has a

natural tendency for storing experiential knowledge and making it available for use. ANNs are among the most effective learning methods to learn and interpret complex real-world sensor data.

The learning methods are classified into three categories i.e. supervised learning,unsupervised learning and reinforced learning. The basic classes of networks namely, single layer feed forward network,multilayer feed forward network and recurrent network.

An ANNs is able to form decision or achieve stable configuration patterns, even if it is given “fuzzy” or incomplete information.The following information’s of ANNs create them valuable and attractive for a forecasting job.

It gives the ability to learn fromexperience and provides a realisticpossible way to solve real worldproblems.

ANNs can generalize and correctly infer the unseen part of the data evenif the sample data contain noisyinformation with least principle.

ANN’s universal functional approximates and provides a continuous function for a desired accuracy.

ANN has the capacity of performing nonlinear modeling without priori knowledge about the relationship between input and output variables. Therefore ANNs are a more general and flexible modeling tool for forecasting.

V. LITERATURE REVIEW

Mayankkumar B Patel,Sunil R Yalamalle [1] examined that prediction provides knowledgeable information regarding the current status of the stock price movement. Stock market data are highly time-variant and are normally in a nonlinear pattern, predicting the future price of a stock is highly challenging. Prediction provides knowledgeable information regarding the current status of the stock price movement. Many different algorithms have been used with neural network. Feedforward MLP(Multi Layer Perceptron) neural network technique is considered to predict the stock price of companies listed under the index of NSE. From the result it can be concluded that MLP (Multi Layer Perceptron)neural network technique gives the satisfactory output Gitansh khirbat ,rahul gupta,sanjay singh,in [2]takes into account factors like Earnings Per Share (EPS) and public confidence and introduces an empirically defined neural network architecture which gives an optimized structure for predicting the future value of a stock by extrapolating the near future value by the present value comparisons. The experimental results obtained after the training and testing of the financial data are very promising. This increase in accuracy of our financial prediction is due to the factors incorporated for forecasting which can give a



clear binary classification for buying or withholding the stock in the current market scenario. Binu Nair, Patturajan, M. ; Mohandas, V. in [3] compares the effectiveness of different types of Adaptive network architectures in one-step ahead prediction of the daily returns of Bombay Stock Exchange Sensitive Index (SENSEX). The performance of each network is evaluated using 17 different performance measures to find the best network architecture. Also, an empirical evaluation of the weak form of Efficient Market Hypothesis (EMH) for the data in reference is carried out here. R.K. and Pawar D.D. in [4] predicated the stockrate because it is a challenging and daunting task tofind out which is more effective and accuratemethod so that a buy or sell signal can be generatedfor given stocks. Predicting stock index withtraditional time series analysis has proven to bedifficult an artificial neural network may be suitablefor the task. Neural network has the ability toextract useful information from large set of data. Inthis paper the author also presented a literaturereview on application of artificial neural network instock market Index prediction. Halbert white in [5] reported some results of anon-going project using neural network modelingand learning techniques to search for and decodenonlinear regularities in asset price movements.Author, focus on case of IBM common stock dailyreturns. Having to deal with the prominent features ofeconomic data highlights the role to be played bystatistical inference and requires modifications tostandard learning techniques that may prove usefulin other contexts. Jing Tao Yao and chew Lim tan in [6] usedartificial neural networks for classification,prediction and recognition. Neural network trainingis an art. Trading based on neural network outputs,or trading strategy is also an art. Authors discuss aseven-step neural network prediction modelbuilding approach in this article. Pre and post dataprocessing/analysis skills, data sampling, trainingcriteria and model recommendation will also becovered in this article. Tiffany Hui-Kuang and Kun-Huang Huarng in[7] used neural network because of theircapabilities in handling nonlinear relationship andalso implement a new fuzzy time series model toimprove forecasting. The fuzzy relationship is usedto forecast the Taiwan stock index. In the neuralnetwork fuzzy time series model where as insampleobservations are used for training and outsampleobservations are used for forecasting. Thedrawback of taking all the degree of membershipfor training and forecasting may affect theperformance of the neural network. To avoid thistake the difference between observations. Thesereduce the range of the universe of discourse. Akinwale adio T, Arogundade O.T and AdekoyaAdebayo F in [8] examined the use of error backpropagation and regression analysis to predict theuntranslated and translated Nigeria Stock MarketPrice (NSMP). The author was used the networktopology to adopt the five input variables.

Thenumber of hidden neurons determined the variables during the network selection. Both theuntranslated and translated statements wereanalyzed and compared. The Performance oftranslated NSMP using regression analysis or errorpropagation was more superior to untranslatedNSMP. David Enke and Suraphan Thawornwong in [9]used machine learning for data mining to evaluatethe predictive relationship of numerous financialand economic variables. Neural network modelused for estimation and classification are thenexamined for their ability to provide an effectiveforecast of future values. A cross-validationtechnique was used to improve the generalizationability of several models. The trading strategiesguided by classification models generate higherrisk-adjusted profits than the buy-and-hold strategyas well as guided by the level-estimation basedforecast of the neural network and regressionmodels. The author decides to deploy the forecastthe stock dividends, transaction costs andindividual-tax brackets to replicate the realisticinvestment practices.

VI. CONCLUSION

Stock market data are highly time-variant and are normally in a nonlinear pattern, predicting the future price of a stock is highly challenging. After studying and analyzing all the papers mentioned in a literature survey and review section we are motivated to study the possibilities of applying other different types of neural network in predicting the share market prices.The future work is to compare the existing algorithms andderive one new ANN model by using clustering and back propagation method of ANN to predict a stock.

REFERENCES

[1] Mayankkumar B Patel,Sunil R Yalamalle ,”Stock Price Prediction Using Artificial Neural Network”, International Journal of Innovative Research in Science,engineering and Technology ,2014.

[2] [2].Gitansh khirbat ,Rahul gupta,Sanjay singh “Optimumneur Network Architecture For Stock Market Forecasting”,IEEE(978-0-7695-4958-3)2013.

[3] [3].Binu Nair, Patturajan, M. ; Mohandas, V.,”Predicting the BSE Sensex: Performance comparison of adaptive linear element, feed forward and time delay neural networks”,2012

[4] [4] Dase R.K. and Pawar D.D., “Application ofArtificial Neural Network for stock marketpredictions: A review of literature”International Journal of Machine Intelligence,ISSN: 0975–2927, Volume 2, Issue 2, 2010,pp-14-17.

[5] [5].Halbert White,” Economic prediction usingneural networks: the case of IBM daily stockreturns” Department of Economics Universityof California, San Diego.

[6] [6].JingTao YAO and Chew Lim TAN ,“Guidelines for Financial Prediction with Artificial neural networks“.

[7] [7].Tiffany Hui-Kuang yu and Kun-Huang Huarng,“A Neural network-based fuzzy time seriesmodel to improve forecasting”, Elsevier, 2010,pp: 3366-3372.



[8] [8].Akinwale Adio T, Arogundade O.T andAdekoya Adebayo F, “ Translated Nigeriastock market price using artificial neuralnetwork for effective prediction. Journal oftheoretical and Applied Informationtechnology , 2009.

[9] [9]. David Enke and Suraphan Thawornwong, “Theuse of data mining and neural networks forforecasting stock market returns, 2005.

[10] [10].Z.H. Khan, T.S. Alin and M.A. Hussain, “Price Prediction of Share Market using Artificial Neural Network (ANN)”, International Journal of Computer Applications, 22, No 2, pp. 42-47, 2011.

[11] [11].H. Al-Qaheri, A. E. Haassanien, and A. Abraham, “Discovering stock price prediction rules using rough sets, Neural network world, vol 18, 181-198, 2008.

[12] [12].Bruce Vanstone & Gavin Finnie, [2009] “An empirical methodology for developing stock market trading systems using artificial neural networks, An international journal of expert systems with applications vol 36, issue 3, 6668-6680.

[13] .Mitra Subrata Kumar [2009] “Optimal combination of trading rules using neural networks” International business research vol 2, no 1 pp 86-99.

[14] .Altay, E., and Satman, M.H. (2005). “Stock market forecasting: Artificial neural networks and linear regression comparison in an emerging market”, Journal of financial management and analysis 18(2), 18-33.

[15] .Senol, D., and Ozturan, M. (2008). “Stock price direction prediction using artificial neural network approach: The case of Turkey,” Journal of artificial Intelligence 1(2), 70-77.

[16] .Yumulu, S. Gurgen, F, Okay N (2004). “A comparison of global, recurrent and smoothed-piecewise neural models for Istanbul stock exchange prediction”, Pattern recognition letters 26(13), 2093-2103.

[17] .Vural B.B. (2007) “financial forecasting with artificial neural networks”, Ankara University, Institute of social sciences, MSc Thesis, Ankara.

[18] .Akel, V., and Bayramoglu, M.F (2008). “Financial predicting with artificial neural networks in crisis periods: The case of ISE 100 index,” International symposium on International Capital Flows and Emerging Markets, Balikesir.

[19] .E.L. de Faria and J.L. Gonzalez (2009); Predicting the Brazilian stock market through neural networks and adaptive exponential smoothing methods, expert systems with applications, Expert system Applications. 36: 12506-12509.

[20] Mahmood Moein Aldin, Hasan Dehghan Dehnavi, Somayye Entezari, (2012). Evaluating the Employment of Technical Indicators in Predicting Stock Price Index Variations Using Artificial Neural Networks (Case Study: Tehran Stock Exchange). International Journal of Business and Management, 7, No.15 pp 25-34.

prediction of stock using artificial neural networks (a survey)

Documents