Integrated Intelligent Research (IIR) International Journal of Web Technology

Volume: 02 Issue: 02 December 2013 Page No.49-55

ISSN: 2278-2419

Forecasting of Stock Prices Using Multi Layer

Perceptron
A. Victor Devadoss1, T. Antony Alphonnse Ligori2
Head and Associate Professor, Department of Mathematics, Loyola College, Chennai, India.
1

2
Ph. D Research Scholar, Department of Mathematics,Loyola College, Chennai, India.
Email : hanivictor@ymail.com, antony_ligori2001@yahoo.com

Abstract - Prediction of stock market has been a challenging model the behavior of stock movements since it can guide their
task and of great interest for researchers as the very fact that financial investments and yield significant profits. Albeit no
stock market is a highly volatile in its behavior. For predicting method has been successful to accurately predict the stock
stock price of Bombay Stock Exchange (BSE), Multilayer price movement till now. ANNs ascertain a hope for the
Networks with dynamic back propagation has been used. The investors, an ability to predict the stock prices.
stock prices are determined and compared with two different
architectures NN1 (3-16-1) and NN2 (3-6-1). Neural Network II. LITERATURE REVIEW
based forecasting of stock prices of selected sectors under
Bombay Stock Exchange show that neural networks have the Artificial Neural Networks are highly flexible function
power to predict prices albeit the volatility in the markets. The approximators that can map any nonlinear function [4] and were
paper is organized as follows. In Section one the volatile nature used initially in the fields of cognitive science and engineering
of stock market is discussed. Section two reviews the literature and later applied in financial time series forecasting. ANNs are
on the applications of ANNs in predicting the stock prices. also being used for a wide variety of tasks in many different
Section three gives an overview of forecasting methods. In fields of business, industry and science. One major application
Section four the concept of Artificial Neural Network area of ANNs is forecasting [5]. ANNs are well suited for
presented. Section five presents the methodology adopted in problems whose solutions require knowledge that is difficult to
forecasting the stock price. In the final section results, future specify but for which there are enough data or observations [6].
direction of the study and conclusion are derived. The first forecasting using ANNs [6] using Wildrow’s adaptive
linear network to weather forecasting. The first study to use the
Keywords -Artificial Neural Network (ANN); Fundamental multi-layer feedforward networks for forecasting purposes is
Analysis; Stock Market; Technical Analysis; Time Series seen in the study [7]. A comprehensive study is done by
Analysis. researchers on the common parameters in designing a
backpropagation neural network and provided step by step
I. INTRODUCTION methodology to design a neural network for forecasting
economic time series data [8].
Prediction of stock market is substantial in finance and is
gaining more attention, due to the fact that if the direction of III. METHODS USED FOR FORECASTING
the market is predicted successfully the investors may be better
guided. Researchers have proposed many models using Let us enumerate some available forecasting methods in
various fundamental, technical and time series forecasting predicting the stock prices.
techniques to give competitive predictions. The existence of
the nonlinearity and volatility of the financial market is A. Fundamental analysis
propounded by many researchers and financial analysts [1]. As
the stock market being dynamic, non-linear, and chaotic in Fundamental analysis is a type of investment analysis adopted
nature it is very difficult to understand because of its volatility, by investors for taking investment decisions and the investors
hence it is of great importance for the investors to know its who follow this approach are called ‘fundamentalists’. They
behavior which would help for their effective investment in it. try to estimate the intrinsic worth of a company’s share, by
Artificial Neural Network (ANN) has the ability to discover studying its sales, earnings, profits, dividends, management
the nonlinear relationship in the input data set without a priori proficiency, and a host of other economic factors that have a
assumption of knowledge of relation between the input and the bearing on the company’s profitability and business prospects.
output [2]. Hence Artificial Neural Networks suits well than The objective is to estimate what the price of a particular
other models in predicting the stock market returns. company’s share out to be and consider this price to be its
Nowadays, stock markets have become an integral part of the intrinsic or true value of the share as it reflects the inherent
global economy. Any fluctuation in the market influences our worth and value. With the help of intrinsic price one can judge
personal and corporate financial lives and the economic health whether the shares are currently over-priced or underpriced in
of a country. Due to its unpredictable behavior there is always the stock market. The fundamentalist makes his money by
some risk to the investment in the stock market [3]. Many buying underpriced shares and later selling them when they
scientific attempts have been conducted on stock market to become over-priced. Fundamental analysis is more useful for
extract some useful patterns and predict their movements. long-term investors.
Moreover, many financial corporations have been in trying to
49
Integrated Intelligent Research (IIR) International Journal of Web Technology
Volume: 02 Issue: 02 December 2013 Page No.49-55
ISSN: 2278-2419
B. Technical analysis
The technical analysis is characterized by a large number of An important application of neural networks is pattern
rules and indicators committed to identify and explain the recognition. Pattern recognition can be implemented by using a
regularity of historical price dynamics. Technical analysis uses feed-forward neural network that has been trained accordingly.
patterns of the price history of a financial instrument in order During training, the network is trained to associate outputs
to provide indications on the future behavior of prices [9]. with input patterns. When the network is used, it identifies the
Technical analysts argue that prices gradually adjust to new input pattern and tries to output the associated output pattern.
information. The Moving Average method (MA) is one of the The power of neural networks comes to life when a pattern that
most used methods of technical analysis. This method has no output associated with it, is given as an input. In this
involves a comparison of the market prices or index with the case, the network gives the output that corresponds to a taught
long MA. The MA method is easy to use and apply in input pattern that is least different from the given pattern. The
investment decision-making or empirical tests [10]. The study of artificial neural networks has been inspired by the
research [11] showed that MA method can generate significant biological learning systems which consist of very complex
forecast value errors and deviations from real prices and is not webs of interconnected neurons. ANNs are built out of
successful in price movement trend generation.Technical densely interconnected units (neurons) where each unit takes a
analysis is commonly used for taking ‘buying’ and ‘selling’ number of real-valued inputs which produces a single real-
decisions in the stock market. This analysis attempts to predict valued output that may in turn be an input to other units.
the future price of a particular share on the basis of a study of ANNs have the ability to learn and thereby acquire knowledge
its price movements in the past. Technical analysts are also and make it available for use. ANNs are among the most
called as ‘chartists’ as they use charts and graphs for keeping a effective learning methods to learn and interpret complex real-
record of share price movements. They believe that an world sensor data [14]. We just recall the notion of neural
elaborate study of share price charts and graphs will reveal network called the Weighted Multi Expert Neural Network
regular and recurrent patterns of price behavior which are (Wt.M.E.N.N) constructed using the fuzzy neural networks.
likely to be repeated in the future. Technical variables most This Wt.M.E.N.N., guarantees equal representation of opinion
frequently cited are shown [12]. They usually ignore all of each expert; hence this method has an advantage over the
fundamental data like sales, earnings, profits, dividends, Fuzzy Neural Networks. Neural Network learning can be either
business prospects of the company, etc. and believe that these supervised one or an unsupervised one. In a supervised
factors have already been taken into account by the market and learning algorithm, learning is guided by specifying, for each
are fully reflected in the current market price of a share. training input pattern the class to which the pattern is supposed
Technical analysis by the very nature of its approach is suitable to belong. In an unsupervised one, the network forms its own
for speculators and short-term traders in shares. classification of patterns. The classification is based on
commonalties in certain features of input pattern. Since the
C. Time series forecasting data is an unsupervised one, we make use of Wt.M.E.N.N. In
Time series forecasting is the analysis of the time series data any supervised learning, a training set of correct input-output
that tries to predict the near future data based on its past data. pairs is given so as to minimize the error, but in an
This is significant in the field of stock market investment, as unsupervised one the output is purely based on the input data.
investors want to make right decisions at right times to We just recall the definition of Neural Network.
maximize their financial profit. Conventional researches used
time series analysis techniques like mixed auto regression A. Definition
moving average (ARMA) and multiple regression models [13].
Time series forecasting usually find a trend in the past data to A neural network is a computational structure that is inspired by
predict future data. The more past data, the easier it is to find a observed process in natural network of biological neurons in the
pattern. However if the history of a stock is short, an accurate brain. It consists of simple computational units, called neurons
analysis and forecast for such little past data is difficult. So, in that are highly interconnected. Each interconnection has a
this case neural networks are described as great tools to use in strength that is expressed by a number referred as weight.
this scenario.
B. Definition
D. Artificial Neural Networks in stock market prediction
Investment in stock market carries a higher risk due to its The bias defines the value of the weighted sum of inputs around
uncertainty and volatility and hence forecasting the stock price which the output of neuron is most sensitive to changes in the
behavior is very difficult. The difficulty arises due to the sum. Now we proceed on to define the notion of Weighted
nonlinear and complex behavior of stock prices. As the Multi Expert Neural Network.In Neural Network bias plays an
primary application of artificial neural networks is in areas important role. So we take the bias as an input with value -1
where problems are ill-defined, data is incomplete or noisy in and its corresponding weight is the sum of the average of the
nature and the environment itself is dynamic. As artificial other input weights. In general, using this newly constructed
neural networks are able to adapt to noisy data and establish Weighted Multi Expert Neural Network (Wt. M.E.N.N.), we
input-output relationship of nonlinear data, the behavior of can extend to `n’ number of experts.
stock price prediction is possible. In the last two decades
extensive researches been attempted through neural networks The class of sigmoid function S, defined by the formula
to forecast stock prices. .

IV. ARTIFICIAL NEURAL NETWORKS

50
Integrated Intelligent Research (IIR) International Journal of Web Technology
Volume: 02 Issue: 02 December 2013 Page No.49-55
ISSN: 2278-2419
Then, the output of the neuron is defined by input vector is applied to the nodes of the network and result of
which becomes a set of outputs for the network at the output
layer. During this phase the weights are all fixed. In the
backward pass, the error term is calculated by finding the
where  is a positive constant (Steepness parameter),  is
difference between actual response of the network and desired
called the bias of the neuron, since  the bias is considered as
response specified to the network and is propagated backward
an input, x0 = –1 and the associated weight w0 = . The through the network. Here the weights are adjusted so as to
quantities Xi and Wi denote the inputs and weights respectively. make the actual response of the network becomes closer to
desired response [15].
C. Neural Network Architectures
The neural network training is an unconstrained nonlinear
An ANN is typically composed of layers of nodes. In the minimization problem in which synaptic weights of a network
popular MLP, all the input nodes are in one input layer, all the are iteratively modified to minimize the overall mean or total
output nodes are distributed into one or more hidden layers in squared error between the desired and actual output values.
between. An MLP is determined by the following variables: The most popularly used backpropagation algorithm is used for
 The number of input nodes training which follows the gradient steepest descent method.
 The number of hidden layers and hidden nodes For the gradient descent algorithm, a step size, called learning
 The number of output nodes. rate must be specified. The learning rate is a constant of
proportionality which determines the size of the weight
The selection of these parameters is basically problem- changes. The weight change of a neuron is proportional to the
dependent. There exists many different approaches such as the impact of the weight from that neuron on the error. A very
pruning algorithm for finding the optimal architecture of an small learning rate requires more training time. One method to
ANN, these methods are usually quite complex in nature and increase the learning rate and thereby speed up training time
are difficult to implement. Furthermore none of these methods without leading to oscillation is to include a momentum term in
can guarantee the optimal solution for all real forecasting the backpropagation learning rule. The momentum term
problems. To date, there is no simple clear-cut method for determines how past weight changes affect current weight
determination of these parameters. Guidelines are either changes. Most neural network software programs provide
heuristic or based on simulations derived from limited default values for learning rate and momentum that typically
experiments. Hence the design of an ANN is more of an art work well. Initial learning rates in the literature are found to
than a science. An artificial neural network is defined as a data vary widely from 0.1 to 0.9. Common practice is to start
processing system consisting of a large number of simple training with a higher learning rate such as 0.7 and decrease as
highly interconnected processing artificial neurons in an training proceeds. Many neural network programs will
architecture inspired by the structure of the cerebral cortex of automatically decrease the learning rate and increase
the brain. There are several classes of neural networks. It is momentum values as convergence is reached.
classified according to the learning mechanisms. The three
broadly classified learning methods are supervised learning, V. APPLICATION OF MLP IN FORECASTING STOCK
unsupervised learning and reinforced learning. There are three PRICE
fundamental classes of networks namely, single layer
Feedforward network, multilayer Feedforward network and Assume that is the data series of stock price, where
recurrent network. and N is the number of data of selected
companies under Bombay Stock Exchange between 1stJanuary
D. Multilayer feedforward network with back propagation 2012and 7thNovember 2013.
algorithm
The following Table I display the selected companies and
MLP is a feedforward neural network with one or more layers sectors used in the study.
between input and output layer. Feedforward means that data
flows in one direction from input to output layer. MLP has Table I. Selected Companies Under Bse For The Present
three layers; an input layer, one more hidden layers and output Study
layer. The input data are fed to the neurons in the input layer
and after processing within the individual neurons of the input
Sl Sector Company
layer the output values are forwarded to neurons in the hidden
No.
layer and finally to the neurons in the output layer. MLPs are
Tata Consultancy
widely used for pattern classification, recognition, prediction 1 Computers-Software
Services Ltd
and approximation.
2 Computers-Software Infosys Technologies Ltd
Connections among the neurons are associated by weights and Healthcare and Dr. Reddy’s Laboratories
3
changing the weights in a specific manner results to learning of Pharmaceuticals Ltd
the associated network. The procedure by which the weight Healthcare and
4 Sun Pharmaceutical Ltd
changes take place in the network is called learning or training Pharmaceuticals
algorithm. The backpropagation algorithm is the most In the present study Multilayer feedforward network (MLP)
commonly used learning technique. The technique consists of with dynamic backpropagation learning has been used. Two
a forward pass and a backward pass. In the forward pass, an networks are constructed namely NN1 and NN2. Both the
network contains an input layer, one hidden layer and one
51
 Integrated Intelligent Research (IIR) International Journal of Web Technology
Volume: 02 Issue: 02 December 2013 Page No.49-55
ISSN: 2278-2419
output layer. The number of neurons in the input layer is three and RMSE can be calculated by using the formula (1), (2) and
which are the three consecutive past prices of the particular (3).
stock under study. The number of hidden layers and the
number of neurons selected in the study has been done --- (1)
heuristically. The hidden layer consists of sixteen neurons in
NN1 and six in NN2 which provide the network with its ability
to generalize. In the study it was started with a fewer number --- (2)
of neurons, but due to the inaccuracy in the initial predictions
the number of neurons were increased in the hidden layer [16].
The number of neurons in the output layer is one as the --- (3)
modeling applied in the study aims to predict one step ahead
closing value in the future forecasting. The sigmoid function is
In this study NEUROPH has been employed to forecast stock
used as a transfer function because it is commonly used for
prices of selected companies and sectors listed in the table1.
time series data as they are nonlinear and continuously
differentiable which are desirable properties for network
Step1: Divide the whole data into two teams 60% for training
training.Table II and Table III show the parameters for the
set and 40% for testing set.
construction of training sets for the companies. In the
following tables LR denotes learning rate and MR denote
Step 2: 60% data has been used to train the network. The
momentum rate.
normalization for the input is done using the formula
Table 2 Training Parameters for NN1
, where x denotes the value that should be
Total Mean
Total Net
Company Name LR MR Squared
Error normalized; denotes the normalized value of x;
Error
Tata Consultancy represents the minimum value of x; represents the
0.4 0.5 0.00035 0.00101
Services Ltd maximum value of x. After the normalization the data (stock
Infosys prices) will be in the range of [0, 1].
0.4 0.4 0.00117 0.00382
Technologies Ltd
Dr. Reddy’s Step 3: A maximum network error of 0.01, the momentum rate
0.2 0.5 0.00036 0.00069
Laboratories Ltd and learning rate range between 0.1 and 0.9 are used in the
Sun training.
0.2 0.3 0.00030 0.00114
Pharmaceutical Ltd
Step4: The trained network obtained in step 3 has been tested
Table 3 Training Parameters For Nn2 with randomly selected data from the 40% testing sets.

To measure the performance of the neural network model used, VI. CONCLUSION AND DIRECTION
Mean Absolute Percentage Error (MAPE), Mean Absolute
Deviation (MAD) and Root Mean Squared Error (RMSE) were The following Table IV displays the Actual value Vs.
calculated. Suppose are actual values and Predicted value against the companies under BSE on randomly
selected dates from the 40% testing set
are the predicted values then the MAPE, MAD
Table V displays the performance of the network adopted in is the preprocessed historic closing prices of the companies.
the study. A highly flexible non-linear modeling technique The model predicts the closing price if three consecutive
ANN has been implemented to forecast the stock prices of trading days’ stock prices are supplied to the model. In a
selected sectors under Bombay Stock Exchange. Two different highly volatile market like Indian Stock Market, if the
neural networks used in this study namely NN1 (3-16-1) and prediction of the direction of the market is possible with fairly
NN2 (3-6-1) where sigmoid function is considered as transfer high accuracy it will guide the investors to reap benefits.
function. The architectures are tested with the testing data set
and the results are predicted. The input data used in the model

Table Iv. Actual Value Vs. Predicted Value Of Stock Prices

Company LR MR Total Net Total Mean
Name Error Actual Squared (NN1-3-16-1) (NN2-3-6-1)
Company Name Date value Error Predicted Forecasting Error Predicted Forecasting Error
Tata 0.5 0.5 0.00032 (Rs.) 0.00045value (%) value (%)
Consultancy 29/4/2013 1370.70 1452.00 5.93 1423.96 3.89
Tata Consultancy
Services Ltd 5/7/2013 1527.35 1604.98 5.08 1551.32 1.57
Services Ltd 0.4
Infosys 0.3 0.00125 0.00330
6/8/2013 1870.50 1787.97 4.41 1742.31 6.85
Technologies
18/2/2013 2776.55 2787.30 0.39 2784.85 0.30
Ltd
Infosys Technologies
8/4/2013 2833.15 2843.44 0.36 2751.08 2.90
Ltd.
Dr. Reddy’s 0.7 0.2 0.00062 0.01033
17/9/2013 3023.35 2909.26 3.77 3348.80 10.76
Laboratories
Ltd 52
Sun 0.7 0.5 0.00018 0.00187
Pharmaceutical
Ltd
Integrated Intelligent Research (IIR) International Journal of Web Technology
Volume: 02 Issue: 02 December 2013 Page No.49-55
ISSN: 2278-2419
6/3/2013 1807.70 1766.09 2.30 1801.16 0.36
Dr. Reddy’s
6/5/2013 1992.75 1934.31 2.93 1950.08 2.14
Laboratories Ltd
6/8/2013 2197.65 2010.44 8.52 2104.98 4.22
Sun Pharmaceutical Ltd 13/03/2013 418.88 420.36 0.35 422.63 0.90
3/4/2013 432.30 403.12 6.75 427.41 1.13
14/6/2013 476.42 478.94 0.53 521.43 9.45

Table V. Evaluation Of The Neural Network

Input
Neural Tata Infosys Dr. Reddy’s Sun
Parameters Forecasting
Network Consultancy Technologies Laboratories Pharmaceutical
(Previous three Performance
Architecture Services Ltd Ltd Ltd Ltd
closing prices)
MAPE 5.14 1.51 4.58 2.54
(NN1-3-16-1) MAD 80.49 45.04 95.62 11.06
RMSE 80.51 66.43 115.68 16.93
MAPE 4.10 4.65 2.24 3.83
(NN2-3-6-1) MAD 68.47 138.61 47.29 17.88
RMSE 81.33 193.84 59.02 26.23

Figure 1. Actual Vs. Predicted stock prices of Tata Figure 3. Actual Vs. Predicted stock prices of Dr. Reddy’s
Consultancy Services Ltd Laboratories Ltd

Figure 2. Actual Vs. Predicted stock prices of Sun Figure 4. Actual Vs. Predicted stock prices of Infosys
Pharmaceutical Ltd Technologies Ltd

Comparing the performance of the two networks, NN1 performs better than NN2 by observing the MAPE results and are shown
in Fig. 5, Fig. 6 and Fig. 7.

53
Integrated Intelligent Research (IIR) International Journal of Web Technology
Volume: 02 Issue: 02 December 2013 Page No.49-55
ISSN: 2278-2419

Figure 5. MAPE comparison for NN1 and NN2

Figure 6. MAD comparison for NN1 and NN2

Figure 7. RMSE comparison for NN1 and NN2

The predicted results show that artificial neural network has Average Comparison”. International Journal of Business, Vol. 8, No.3, pp.
267-284, 2003.
been able to predict stock prices and ensure that it is suitable [11]A. Dzikevicius, S. Saranda, and A. Kravcionok, “The Accuracy of
for forecasting with better accuracies. Limitations and Future Simple Trading Rules in Stock Markets”, Economic and Management, No.15,
Scope of Improvement In this study only the historic prices of pp. 910-916, 2010.
stock has been used for the prediction problem with artificial [12]B. Vanstone and G. Finnie, “Combining Technical Analysis and Neural
Networks in the Australian Stockmarket”. Bond University
neural network. In order to improve the accuracy of the model, ePublications@bond. Information technology papers, 2006.
macroeconomic factors and international stock market data can [13]W. Huang, Y. Nakamori and Y. Wang, “Forecasting stock market
also be used as input variables. Technical analysis indicators movement direction with support vector machine. Computers and Operations
can also be used in the input variables and can be checked for Research”, No. 32, pp. 2513-2252, 2005.
[14]T. Mitchell, “Artificial Neural Networks: Based on Machine Learning”,
improvement in the performance of the network. Mc Graw Hill, 1994.
[15]Simon Haykin, “Neural Network A comprehensive Foundation”, Second
REFERENCES edition, Prentice Hall, pp. 161-173, 1998.
[1]A. Abhyankar, L.S. Copeland and W. Wong, “Uncovering nonlinear [16]S. Chaigusin, C. Chirathamjaree and J.M. Clayden, “The Use of Neural
structure in real-time stock market indexes: The S&P 500, the DAX, the Networks in the Prediction of Stock Exchange of Thailand (SET) Index”,
Nikkei 225 and the FTSE-100”, Journal of Business & Economic Statistics, No Proceedings of International Conference on Computational Intelligence for
15, pp 1-14, 1997. Modelling, Control and Automation, Vienna, Austria, pp. 670-673., 2008.
[2]M.T. Hagan, H.B. Demuth and., M. Beale, “Neural network design”. PWS [17]S. Saigal and D. Mehrotra, “Performance Comparison of Time Series Data
Publishing Company, Boston, 1996. Using Predictive Data Mining Techniques”, Advances in Information Mining,
[3]M.R. Hassan and B. Nath,, “Stock Market Forecasting using Hidden No.1, pp. 57-66, 2012.
Markov Model: A new approach”, Proceedings of 5 th International conference [18]http://neuroph.sourceforge.net
on Intelligent systems Design and Applications, 2005. [19]http://in.finance.yahoo.com
[4]H. White, “Learning in neural networks: A statistical perspective”, Neural [21]M.J.C. Hu,. “Application of the adaline system to weather forecasting”.
Computation. No. 4, pp. 425-464, 1989. Master thesis, Technical Report 6775-1. Stanford Electronic Laboratories,
[5]R. L. Wilson and R. Sharda, :Bankruptcy Prediction Using Neural Stanford, CA, June, 1964.
Networks”, Decision Support Systems, No. 11, pp. 545-557, 1994. [22]W. Huang, Y. Nakamori and Y. Wang, “Forecasting stock market
[6]G. Zang, B.E. Patuwo and M.Y. Hu, “Forecasting with artificial neural movement direction with support vector machine. Computers and Operations
networks: The state of the art”, International Journal of Forecasting, 14, pp. 35- Research”, No. 32, pp. 2513-2252, 2005.
62, 1998. [23]Z.H. Khan, T.S. Alin and M.A. Hussain, “Price Prediction of Share Market
[7]A. Lapede and R. Farber, “Nonlinear signal processing using neural using Artificial Neural Network (ANN)”, International Journal of Computer
networks: prediction and system modeling”, Technical Report LA-UR-87- Applications, 22, No 2, pp. 42-47, 2011.
2662, Los Alamos National Laboratory, Los Alamos, NM, 1987. [24]M. Majumder and M.A. Hussain, “A. Forecasting of Indian Stock Market
[8]M. Kaastgra and M. Boyd, “Designing a neural network for forecasting Index using Artificial Neural Network” , unpublished article.
financial and economic time series”, Neurocomputing, No. 10, pp. 215-236, [25]S. Rajasekaran and G.A. Vijayalakshmi, “Neural Networks, Fuzzy logic
1996. and Genetic Algorithms”, PHI, 2003.
[9]M. Pring, “Martin Pring’s Introduction to Technical Analysis”, Singapore: [26]R. Lawrence, “Using Neural Networks to Forecast Stock Market Price",
McGraw-Hill, 1999. 1997.
[10]U. BenZion, P. Klein, Y. Shachmurove and J. Yagil, “Efficiency
Differences Between the S&P 500 and the Tel-Aviv 25 indices: A Moving
54