2017 International Conference on Technical Advancements in Computers and Communications
Developing a Prediction Model for Stock Analysis
R. Yamini Nivetha
Department of Computer Science and Engineering.
Adhiparasakthi Engineering College
Melmaruvathur, Kanchipuram District
nive.yamini91@gmail.com
Abstract—Prediction in the stock market is challenging and
complicated for investors. Many researches have performed to
sense the future market movements. In the stock market, social
media data have high impact today than ever. In this work,
various prediction algorithms are analyzed to build a prediction
model. The prediction model will be based on monthly prediction
and daily prediction to forecast the next day market price. This
model estimates the open value of the next day in the market.
Sentiment Analysis needs to identify and extract sentiments from
each individual in the social media. The correlation between the
sentiments and the stock value is to be determined. A
comparative study of these three algorithms: Multiple Linear
Regression, Support Vector Machine and Artificial Neural
Network are done. The stock price is predicted by sentiment
analysis with the best forecasting algorithm.
Keywords—Stock Market; Sentiment Analysis; Multiple Linear
Regression (MLR); Support Vector Machine (SVM); Artificial
Neural Network (ANN).
I. INTRODUCTION
The stock market is always one of the most popular
investments due to its high profit. As the level of investing and
trading grew, people searched for tools and methods that would
increase their gains while minimizing the risk. Most of the
trading in Indian Stock Market takes place on its two stock
exchanges: the Bombay Stock Exchange (BSE) and the
National Stock Exchange (NSE). The two prominent Indian
Market Indexes are Sensex and Nifty. Stock market prediction
is complicated since the prices in the stock market are dynamic
in nature. The prominent input factors of the company that
have an impact on prediction are Price-To-Earnings Ratio (PE),
Debt-To-Equity Ratio and Price Flow Ratio (PF).
In this paper, to make a comparative study of stock
predicting algorithms and determine the best predicting
algorithm. The Linear Regression which contains more than
two predictor variables is Multiple Linear Regression. It
measures the relationship between the two variables. The least
square value is calculated to obtain the prediction results. The
support vector machine is a supervised learning model to
classify the training data. The hyper-plane is identified and the
margin of the training data is maximized by Optimal
Separating hyper-plane. The support vectors which are closest
to the hyper-plane and it is in equal distance from both the
classes are identified. The value of support vector is evaluated
by assigning the three weights to be learned from the SVM
978-1-5090-4797-0/17 $31.00 © 2017 IEEE
DOI 10.1109/ICTACC.2017.11
Dr. C. Dhaya
Department of Computer Science and Engineering.
Adhiparasakthi Engineering College
Melmaruvathur, Kanchipuram District
dhaya.c@gmail.com
model. The Artificial Neural Network is a deep learning
approach in hidden data neurons, to forecast the price of a
stock. A Neural Network consists of an input layer, hidden
layer and output layer. The weights are assigned for each input
layer. The summation function is calculated for the weighted
output. The output layer calculates the sigmoid function for the
summation value. The final output is the predicted stock price.
The prediction model predicts daily prediction and monthly
prediction. Sentiment Analysis is combined with a best
predicting algorithm to refine the outcomes in the stock market.
Hence the prediction model acts as a stock market broker in
finance and business streams.
II. RELATED RESEARCH
D. Kumar and S. Murugan have introduced a new method
by combining time-series data with Artificial Neural Network
(ANN) [1]. This prediction model is based on a feed-forward
ANN with back propagation. The performance of the
prediction model is analyzed using some factors. The key
factor includes Mean Absolute Error (MAE), Mean Absolute
Percentage Error (MAPE), Percentage Mean Absolute
Deviation (PMAD) and Mean Squared Error (MSE). This
performance is calculated by BSE100 and NIFTY MIDCAP50.
K. Abhishek, A. Khairwa, T. Pratap and S. Prakash have
developed a forecasting model for the Microsoft Corporation.
This forecasting model includes a two-step procedure [2]. In
first-step, ANN go through training with the genetic algorithm.
This algorithm teaches the prediction model to identify the new
trends. In a second step the results are obtained by analyzing
the trained prediction model to the dataset.
M. Hagenan, M. Liebmann, M. Hedwig and D. Neumann
have experimented with the dataset of NSE and news data from
social media [3]. They designed the word combination
technique with German adhoc messages and stock prices. The
text mining is evaluated by Chi-Based feature selection. The
Support Vector Machine algorithm is used to predict the stock
price. The classification accuracy is boosted by feedback based
feature selection.
Arti Buche and Dr. M. B. Chandak discuss the text opining
mining in stock market reviews of Economic Times [4]. This
method deals with the polarity of the text in two processes,
namely POS tagger and SentiWordNet. This method is
1
implemented by Natural processing language and statistical
parameter. This paper has surveyed on different methods to
highlight the factors that have an impact on market price.
Z. Khan, T. Alin and Md. A. Hussain highlighted the
consequences in prediction of stock value [5]. The system is
incorporated by backward propagation with artificial neural
network. The general index (GI), net asset value (NAV),
earnings per share (EPS) and volume of the market may affect
the market movement. The system with this algorithm is
experimented in BSE stock market.
III. PROBLEM DEFINITION
The stock market prediction is difficult since the stock price
is dynamic in nature. To reduce the false forecasts of the stock
market and increase the ability to predict the market
movements. To avoid the risk and the complex in predicting
stock price.
IV. SYSTEM DESIGN
Fig. 1. Stock Prediction Model
The stock price is predicted in the prediction model by
using the historical data and text opinion mining in social
media data. The Yahoo finance data contains the historical data
of three months of various companies. It is used for monthly
prediction. The Economic Times data and Twitter data contains
the reviews (or) news which is expressed by each individual is
evaluated by the sentiment value. This data is used for daily
prediction. The Sentiment Analysis is combined with the best
forecasting algorithm.
V. ALGORITHMS
A. Data Collection and Processing
The Yahoo finance data contain BSE data and NSE data.
The financial index of BSE is SENSEX and for NSE is NIFTY.
This finance data contain many attributes of the stock market.
Some of the attributes are opening value, closing value,
adjacent close value, min value, max value, volume, dividend,
date and time of the day. The key attributes are selected for
stock market prediction in this work. The collected data are
2
processed by normalization and used as the input data. For
monthly prediction and daily prediction historical data from
Yahoo finance is used.
B. Multiple Linear Regression
Regression is a data mining task of predicting the stock
price by implementing a model based on one or more
attributes. This technique is a compact method of mathematical
representation. It represents the relationship between the
response parameter and the input parameter in a given
prediction model. Prediction of the outcome y from the input
parameter x of the form
y = β0 + β1 x1 + β2 x2… + βk xk (1)
y – Outcome, β0 – Intercept, β1, β2, βk – Partial Regression
Co-efficient, x1, x2, xk – Input Parameters.
The least square value is determined by the current market
price y from the PE ratio and mean of the input parameters
y = a + bx (2)
y – Current market price, x – Price-to-Earnings ratio (PE
ratio), a – Mean of y – b (Mean of x), b – Cov(x, y).
C. Support Vector Machine
In machine learning, Support vector machine is supervised
learning models with associated learning algorithms to identify
the data. This process includes the classification and regression
methods to analyse the prediction. It implements the margins
adoption to predict the market price. The support vector means
the closest hyper-plane and it is equal in distance are identified.
The value of support vector y is calculated using the two
attribute case
y = w0 + w1 x1 + w2 x2 (3)
y – Outcome, w0, w1, w2 – three weights to be learned by
SVM model, x1, x2 – input parameters.
Fig. 2. SVM Model
D. Artificial Neural Networks
The artificial neural network is deep learning at hidden
layer neuron. Artificial Intelligence classifiers are used to train
the prediction model to identify the trends in a complicated
environment. The input value is assigned with different weights
in the input layer. The output of the input layer is fed into the
hidden layer neuron as the input. The summation function is
evaluated in the hidden layer. It learns every prediction at each
node. If any hidden layer neuron is not completed, then it is fed
into the summation function.
h = ∑ wi.xi (4)
h – Summation function, ∑wi – Weight of the input, xi –
Input value.
The output of the hidden layer is given as input to the
output layer where the sigmoid function O is determined using
the summation function h.
O = 1 / 1+e (-h) (5)
Fig. 3. Process of ANN
E. Sentiment Analysis
The expression of each individual varies and it is analysed
in social media data. The polarity of each word is evaluated.
The polarity can be neutral, positive and negative. The
correlation between the news and actual price is determined.
The Sentiment value in a row S is initial zero. For each word in
a row do the Sentiment Calculation.
Sent_Word (word, dictionary) (6)
Sent_Word – compares the word with the dictionary.
S = Sent_Word (7)
S – Calculates the value of Sentiment Analysis of the data.
VI. TRADING MODEL
A. Choosing the Best Algorithm
To analyse the best prediction algorithms in stock market
various experiments are examined. Many key factors of stock
market are considered to forecasts the stock price. The factors
are opening price, closing price, adjacent close, volume and
dividend of the stock. The performance is measured for the
analysed algorithms. The best algorithm which forecasts the
results is found and the further design is implemented. The
results exhibit that Artificial Neural Network is the best
predicting algorithm among the three algorithms in stock
market.
B. Proposing a Prediciton Model
The prediction model is designed to forecast the monthly
and daily prediction in stock market. The historical data and
news from social data is collected to design the prediction
model. The monthly prediction uses the historical data of NSE
and BSE data. The daily prediction uses the historical data and
the social media data which includes Economic Times
Website, Money Control Website and Twitter. The expression
of each individual varies and it is analysed in social media data.
The expression of each individual varies and it is analysed in
social media data. The polarity of each word is evaluated. The
polarity can be neutral, positive and negative. The correlation
between the news and actual price is determined.
VII. CONCLUSION
The purpose of this study is to compare the performance of
the three prediction algorithms Multiple Linear Regression,
Support Vector Machine, Artificial Neural Network in the
stock market. The Multiple Linear Regression algorithm is less
developed state which measures the relationship between the
stock price and volume. The Support Vector Machine
algorithm is a two-class classifier for the learning model. The
Artificial Neural Network is the classification algorithm for
deep learning. The result exhibits that the deep learning
algorithm performs better than the MLR and SVM. In deep
learning algorithm the hidden layer neuron learns in every
prediction. Hence the output layer neuron produces the best
outcome. Artificial Neural Network is the best predicting
algorithm.
REFERENCES
[1] Kumar, D. Ashok, and S. Murugan. "Performance analysis of Indian
stock market index using neural network time series model." In Pattern
Recognition, Informatics and Mobile Engineering (PRIME), 2013
International Conference on, pp. 72-78. IEEE, 2013.
[2] Abhishek, Kumar, Anshul Khairwa, Tej Pratap, and Surya Prakash. "A
stock market prediction model using Artificial Neural Network."
In Computing Communication & Networking Technologies (ICCCNT),
2012 Third International Conference on, pp. 1-5. IEEE, 2012.
[3] Hagenau, Michael, Michael Liebmann, Markus Hedwig, and Dirk
Neumann. "Automated news reading: Stock price prediction based on
financial news using context-specific features." In System Science
(HICSS), 2012 45th Hawaii International Conference on, pp. 1040-1049.
IEEE, 2012.
[4] Buche, Arti, M. B. Chandak, and Akshay Zadgaonkar. "An Approach
for Online Analysis using Expectation Maximization." International
Journal of Innovative Research in Computer & Communication
Engineering 1 (2013).
[5] Khan, Zabir Haider, Tasnim Sharmin Alin, and Md Akter Hussain.
"Price prediction of share market using artificial neural network
(ANN)." International Journal of Computer Applications 22, no. 2
(2011): 42-47.