Research Paper Computer Science E-ISSN No : 2454-9916 | Volume : 2 | Issue : 3 | March 2016

SHARE
MARKET
PREDICTION
USING
ARTIFICIAL
NEURAL

NETWORK

1 1 1 1
Swanand S. Sathe | Shalaka M. Purandare | Pallavi D. Pujari | Sagar D. Sawant
1
TSSM's BSCOER Department of Computer Engineering.

ABSTRACT
The share market is dynamic in nature that means to predict share price is very complex process by general prediction or computation method. Its main reason is that
there is no linear relationship between the market parameters and target closing price, so use of Neural Network is a choice of interest for share market prediction. Apart
from Neural Network the proposed system consists of three more algorithms: Regression, Forecasting and Moving average. Using these algorithms, the basic idea
behind this system is to predict the future share values based of the past values. Also, use of these algorithms leads the system to result in better output as well as
increases the efficiency.

KEYWORDS: Share, BSE, Sensex, Closing price, Past data.

I. Introduction data, need of ANN, importance of stock indices, survey of the previous works
Artificial Neural Network plays important role in time series analysis and and it investigates neural network models for time series in forecasting.
prediction. Share market prediction is one of complicated problem because it is
dynamic in nature i.e. sensex value is consistently changing. Hence it is very D. Forecasting of Indian stock market using time-series ARIMA Model
difficult to identify the patterns and predict the share value. Existing system are (2014):
not well efficient because of dynamic nature of share market. In this paper we are According to this paper an application of ARIMA model based on which we
trying to improve accuracy of share market prediction using four different predict the future stock indices which have a strong influence on the performance
algorithms as Neural network, Regression, Forecasting and Moving Average. of the Indian economy. The Indian Stock market is the centre of interest for many
Using this four algorithms we will try to improve accuracy of existing system economists, investors and researchers and hence it is quite important for them to
from 20%-30% to 50%-60%. This will help share broker and customer to have a clear understanding of the present status of the market. To establish the
purchase or sell the share and to invest money. model author applied the validation technique with the observed data of sensex
of 2013.
II. Literature survey
In the Literature survey we are analyzing four papers which contains different III. Existing System
methods or models like moving average, Forecasting, Neural network and Stock market prediction is an act to determine future stock value (share price).
Regression algorithm. We are trying to cover all these methods or algorithms to This prediction takes place by taking the past share values in to consideration.
obtain better accuracy than existing systems. For this the existing system makes use of algorithms such as ANN (Artificial
Neural Network), ARIMA model, Time series prediction etc. Efficiency of these
A. Artificial Neural Networks for Forecasting Stock Price (2008): algorithms is less as compare to the proposed system algorithm. There is no such
According to This paper The objective is to be able to develop a long term pricing a system which makes use of four algorithms in one system. Hence that leads the
relationship between stocks and profit. Statistical arbitrage strategies have existing systems to be less efficient.
always been popular since the advent of algorithmic trading. In particular,
Exchange traded fund (E.T.F.) arbitrage has attracted much attention. Trading IV. Proposed System
houses have tried to replicate ETF arbitrage to other stocks. Thus, the objective is We use artificial neural network methods like Forecasting, Linear regression,
to be able to develop a long term pricing relationship between stocks and profit and Moving averages. In forecasting method the system is taking the three days
from their divergence from this relationship. In this paper, we have developed a last and the current year stock portfolio closing price from the predicted date and
feasible trading strategy on this concept. Artificial neural networks have been performs calculations on it for predicting the stock portfolio price. Moving
deployed to model the pricing relationship between elements in a sector. All averages method, system is take the ten days stock portfolio closing price form
prices have been considered at the same instant, thereby allowing us to make the predicting date and calculate the stock price.
trading decisions in accordance with our predictions. Supervised learning
algorithms have been used to train the network.

B. Stock Market Prediction Using Artificial Neural Networks (2012) :

According to this paper the authors, the aim of this project is implementation of
neural networks with back propagation algorithm for stock market. Borrowing
from biology, researchers are exploring neural networks - a new, non algorithmic
approach to information processing. A neural network is a powerful data-
modeling tool that is able to capture and represent complex input/output
relationships. The motivation for the development of neural network technology
stemmed from the desire to develop an artificial system that could perform
“intelligent" tasks similar to those performed by the human brain.

C. Performance Analysis of Indian Stock Market Index using Neural

Network Time Series Model (2013):
According to this paper, A time series is a collection of observations made
chronologically. The nature of time series data includes: large in data size, high
dimensionality and necessary to update continuously. Forecasting based on time
series data for stock prices, currency exchange rate, price indices, etc., is one of
the active research areas in many field viz., finance, mathematics, physics,
machine learning, etc. Initially, the problem of financial time sequences analysis
and prediction are solved by many statistical models. During the past few
decades, a large number of neural network models have been proposed to solve
the problem of financial data and to obtain accurate prediction result. The
statistical model integrated with ANN (Hybrid model) has given better result
than using single model. This work discusses some basic ideas of time series Figure 1. System Architecture
Copyright© 2015, IERJ. This open-access article is published under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License which permits Share (copy and redistribute the material in any
medium or format) and Adapt (remix, transform, and build upon the material) under the Attribution-NonCommercial terms.

International Education & Research Journal [IERJ] 74

 Research Paper E-ISSN No : 2454-9916 | Volume : 2 | Issue : 3 | March 2016
In this method a alpha is set for the predicting value it is a constant for this 3. Then subtract the Linear Regression from set
method. Regression method, it is a statistical assessing the association between ytmp = lr.subtractLinearRegression(null, ytmp), Here, ytmp is the copy of
two variables. It is used to find relationship between two variables. Neural original set
networks method has effective, general purpose approach for pattern
recognition, classification, and clustering and especially time series prediction 4. Apply the fourier analysis on the set
with a great degree of accuracy. Nevertheless, their performance is not always interval1 = fa.transform(ytmp);
satisfactory. Back propagation algorithm is the best algorithm to be used in Feed
forward neural network because it reduces an error between the actual output and 5. After processing on data, normalize the data by dividing by constant value
desired output in a gradient descent manner.
6. Creates new Multi LayerPerceptron with specified number of neurons in
V. Algorithms layers
A. Moving Average Algorithm MultiLayerPerceptron(TransferFunctionTypetransferFunctionType,int..neuron
Get the last n records stock value from predicted date sInLayers)

1. Calculate Value for add 7. Provides implementations of specific neural network learning algorithm
add=actualValues[0]+(5*(actualValues[0]/100)); MomentumBackpropagation learningRule =((MomentumBackpropagation)
neuralNet.getLearningRule())
2. Calculate Pre-Assume value
preassume=actualValues[0]+add 8. Set learning algorithm for this network
public void setLearningRule(LlearningRule)
3. Set static value alpha
alpha= 0.7 9. Performs calculation on whole network
public void calculate()
4. Set one fix predicted value
10. Returns network output vector
5. predictedvalues[0]=preassume+(alpha*(actualValues[0]-preassume)); public double[] getOutput(),Output vector is an array collection of Double
values
6. Apply the formula for Prediction
predictedvalues[i+1]=predictedvalues[i]+(alpha*(actualValues[i+1]- VI. Advantages
predictedvalues[i])); 1. Dynamic in nature.
2. High Accuracy.
B. Regression Algorithm 3. Noise Tolerance.
1. Take a past few close values from any company. That is yi. and take constant 4. Ease of maintenance.
values that is xi 5. Share broker can increase his/her and customer's profit by predicting stock
value.
2. Calculate the slop i.e. b = Σ [ (xi - x)(yi - y) ] / Σ [ (xi - x)2]
VII. Disadvantages
3. Calculate intercept i.e. a = y - b1 * x 1. Problem in updating of data.
2. Previous systems cannot predict the share market values efficiently
4. Now calculate (Y)=a+bX and here we got value of dependent variable.
REFERENCES
5. Calculate the coefficient of determination. That is R2 = { ( 1 / N ) * Σ [ (xi - x) * [1] Wenping Zhang, Chunping Li, Yunming Ye, Wenjie Li and Eric W.T. Ngai, "Dynamic
(yi - y) ] / (σx * σy ) }2 Business Network Analysis for Correlated Stock Price Movement Prediction.", 1541-
1672/15/31.00 © 2015 IEEE.
6. For calculating coefficient of determination we need σx , σy . [2] Debadrita Banerjee," Forecasting of Indian Stock Market using Time-series ARIMA
Model",2014 2nd International Conference on Business and Information Management
2 2
7. So firstly calculate σx = sqrt [ Σ ( xi - x ) / N ] then σy = sqrt [ Σ ( yi - y ) / N ]. (ICBIM).
[3] D. Ashok kumar, S. Murugan, " Performance Analysis of Indian Stock Market Index
8. Coefficient of determination gives us variation of dependant variable and using Neural Network Time Series Model",Proceedings of the 2013 International
relation among the dependant variable and independent variable. Conference on Pattern Recognition, Informatics and Mobile Engineering (PRIME)
February 21-22.
[4] Prakash Ramani, Dr.P.D.Murarka,”Stock market Prediction Using Artificial Neural
C. Forecasting Algorithm Network”, International Journal of Advanced Research in Computer Science and
1. Forecast calculation Software Engineering, volume 3 issue 4, April 2013.
Take the last & current year stock price. [5] Neelama Budhani, Dr.C.K.Jha, Sandeep K. Budhani “Stock market prediction using
Eg. Last year = L1,L2,L3.. artificial neural network ”, International Journal Of Computer science And
Current year = C1,C2,C3.. Engineering Technology, volume 3 no.4 April 2012.
[6] K. K. Sureshkumar, Dr. N. M. Elango, Performance analysis of Stock price prediction
2. Calculate summation of stock prices. using Artificial neural Networks,Global journal of computer science and Technology,
Sum1=L1+L2+L3 volume 2 issue 1 version 1.0 January 2012.
Sum2=C1+C2+C3 [7] Zabir Haider khan, Tasnim Sharmin Alin md. Akter Hussain,price prediction of share
market using Artificial Neural Network(ANN), International Journal of computer
3. Calculate pf. application(09758887) volume 22 no.2, May 2011.
pf =(Sum1/Sum2) [8] K. K. Sureshkumar, Dr. N. M. Elango ,An Efficient Approach to forecast Indian Stock
Market Price and their Performance analysis,International journal of computer
4. Calculate sn using formula 1. application (09758887) volume 34, no 5, November 2011.
sn=n1+n2+n3
[10] Zhou Yixin Jie Zhang, Stock data analysis based on BP neural network, 2010 second
Here n1,n2,n3 are stock values of current year multiplied by pf. international conference on communication Software and Network.

5. Finally calculate Percent Of Accuracy [11] Ramnik Arora (Y5365),"artificial neural networks for forecasting stock price
",Submitted to : Dr. S.K. Mathur Course : Eco 311,2008.
POA = sn / pn * 100 where pn is summation of current year stock value.

6. If POA > 105 then add alpha value Else subtract alpha value.
Where alpha value = 0.7

D. Neural Network Algorithm

1. Get the stock data training set

2. Calculate the Linear Regression factor on stock closing value

lrFactor=(rfactor / y.length) * 100.0D
rfactor = rfactor + Math.abs(diff / y[j]), Here, y[j] is the stock close values set
diff = y[j] - (aX + bX * x[j])
bX = sum1 / sum2
aX = ym - bX * xm

75 International Education & Research Journal [IERJ]