www.ijcrt.

org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882

Share Market Analysis and Prediction

MAYURKUMAR BORSE MANOJ CHAVAN
CHETAN BHAGAT
Department of Information Department of Information
Department of Information
Technology Technology
Technology
G.H.Raisoni(GHRCEM) G.H.Raisoni(GHRCEM)
G.H.Raisoni(GHRCEM)
Pune, India Pune, India
Pune, India

DNYANESHWAR JADHAV MS. MINAXI DOORWAR

Department of Information Department of Information
Technology Technology
G.H.Raisoni(GHRCEM) G.H.Raisoni(GHRCEM)
Pune, India Pune, India

Abstract—Nowadays, the prediction of share market prices The attributes that makes a prediction model depends upon
and conditions has become a major researched topic amongst the factor on which market performance can be depending.
the data scientists, investment bankers, and stock brokers. As, Different types of methods have been developed to forecast
the behavior of share market is very nonlinear and volatile in the behavior of stock market prices based on previous
nature, it makes a very high-risk investment. Consequently, a historical data. As per recent survey Share prices prediction
lot of researchers have came up with their efforts to forecast with conventional statistical methods has proven to be less
the share market and average movement. Researchers have effective because of nonlinear characteristics of financial
used various methods in computer science and economics in time series. On the contrary, machine learning-based
their illustrate to gain a piece of this volatile information and methods, such as Artificial Neural Networks (ANNs),
make great fortune out of the share market investment. The support vector machine (SVM) ,linear regression, recent
approaches like data mining and machine learning trends, LSTM model and data mining system offer useful
approaches can incorporate into Business Intelligence (BI) tools for forecasting noise environments like share market.
systems to help users for decision support in many real-life
applications. This paper presents the brief survey of II. RELATED WORK
application of machine learning in share market prediction A. Stock/Share Market
and investigates various techniques for the share market
A place where shares of public listed companies are traded is
prediction using like Artificial Neural Network (ANN) and
known as Stock Market. It is a regulated financial and well
Support Vector Machine (SVM). ANN is non- linear and
organized market. In stock market, securities such as bonds,
non- parametric classifier which is viable for forecasting of
notes and shares are bought and sold at price and this prices
share prices. Support Vector Machine focuses on marginal
governed by the forces of demand and supply. Stocks are
values rather than average values for the classification
bunches of shares. Stock market serves as, market where
predicting model. The aim of this paper is to provide a review
corporations, governments, municipalities and other
of the application of machine learning in share market
incorporated bodies can raise their capital by taking saving of
prediction to determine what can be done in the future.
the investors into productive ventures. Also, Share market is
Keywords—Share Market, Machine Learning
a place where stock investor can sell their securities to other
I. INTRODUCTION stock investors for cash. And this reduces the risk of
investment and maintaining liquidity in the system.
Predicting the future of share price has always been an
adventurous and attractive task for the investors individuals. B. Machine Learning Algorithms
This kind of prediction becomes more fascinating when it
This section continues Introduction to machine learning and
involves money and risk like predicting share market. Goal
evaluation of the machine learning techniques. Some of the
behind making any financial investment is to achieve above
algorithms we had considered are as follows:
average return for invested money while maintaining certain
level of involved risks but as the share market is a very Computer algorithms Studies that improve automatically
complex, volatile and non-linear dynamical system, share through experience and by the use of data is known as
market prediction has become a tough challenge for Machine learning. It is part of Artificial Intilligence. This
researchers and investors. Lot of Research has been done on concept focuses on the computer development programs that
share market prediction by researchers of different fields can access given data and use it learn for themselves. The
including the business and computer science. Researchers concept start with observing the data, such as examples,
have tried different approaches for share market prediction. direct experience, or instruction, in order to look for patterns
in data and make better decisions in the future based on the
sample data that we provide. The main goal of machine

IJCRTI020009 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 35

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882

learning is to allow the computers learn automatically without es to approach for monetary newspaper
human interaction or assistance and adjust actions Analyz article analysis exploitation many
accordingly. Machine learning algorithms were categorized e Stock completely
as unsupervised and supervised. Market different matter representations:
s Bag of Words, Noun Phrases, and
Flask Framework
Named Entities. Through this
Flask is written in Python and it is a micro web framework. approach, they investigated an
It is classified as a micro-framework because Flask does not oversized range of economic news
require particular tools or libraries that's why it is classified articles and stock quotes covering
as micro-framework. Flask does not have database stocks listed on the
abstraction layer form validation or any other components S&P500 throughout a five-
where pre-existing third-party libraries provide common week amount from Oct twenty six,
functions. 2005 to Gregorian calendar
month twenty-eight, 2005.
.COMPARISON OF LITERATURE They calculable a separate stock v
S Publi Metho Description alue twenty
r shed dology minutes when a newspaper
n Year article was free. exploitation AN S
o VM by-product specially tailored
1 (200 Studies Within the planned learning for separate numeric prediction,
8) [1] Using paradigm, a genetic rule is 1st and models containing completely
Geneti different stock-specific variables,
want to choose input options for
c they showed that the model
LSSVM learning. Then another
Algorit containing each article terms and
GA is employed for parameter stock value at the time of
hms improvement for the LSSVM.
with article unleash provided
Finally, the evolving LSSVM the nearest estimate to the
Other learning paradigm with the
Techni particular future
simplest feature set, optimum stock value, identical direction
ques to
parameters, and a mixed kernel is of value movement because
Analyz
employed to predict stock the future value, and therefore
e
Stock exchange movement direction in the highest come back employing
Market terms of historical knowledge a simulated trading engine.
s series. For analysis functions,
testing is completed mistreatment 3 (200 A dynamic fuzzy model
knowledge from 3 stock indices – 9) Studies is projected by Chiu and Chen in
the S&P five hundred, stock Using combination with a SVM to
index Industrial Average, and Geneti explore stock exchange dynamics.
also the big apple stock market c The fuzzy model integrates input
Index. the whole knowledge set Algorit variables exploitation factors
of monthly values covers the hms with variable degrees of influence.
amount from Gregorian calendar with A GA adjusts
Other the prestigious degree of
month 1926 to Gregorian
Techni every input variable dynamically.
calendar month 2005 with a
ques to The SVM is then wont
complete of 960 observations.
Analyz to predict exchange dynamics. A
Experimental results reveal that e multi-period experiment is
the planned evolving LSSVM Stock meant to
will manufacture some prediction Market simulate exchange volatility.
models that area unit a lot of s The sixty-one input
simply taken as a result of they variables within
use a smaller range of prophetic the study embody exchange techni
options and area unit a lot of cal indicators, futures
economical than different exchange technical indicators,
parameter improvement and economic science variables. to
strategies. judge the performance of the new
2 (200 Studies A unique study by Schumaker and integrated model, they compare it
9) Using Chen used an SVM in conjunction with ancient forecast ways.
Suppor with matter analysis observing the 4 (201 Studies Yeh, Huang and Lee
t impact of stories articles on 1) Using address issues that
Vector stock costs. They developed Suppor arise once mistreatment support
Machin a prognosticative machine learning t vector regression to forecast stock

IJCRTI020009 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 36

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882

Vector exchange values once handling ker CNX smashing (50 of the most
Machin nel operate hyperparameters. important stocks found on the
es to Typically, a hyperparameter may National exchange (NSE) of
Analyz be a parameter whose price is India) and therefore
e Stock ready before the the S&P Mumbai exchange (BSE)
Market training method begins. In their Sensex for the amount Jan 2003
s system, blessings from totally to December 2012.
different hyperparameter
settings will be combined and 7 (201 Studies Dash and Dash introduce a
overall system performance will 6) Using unique decision network using
be improved. They develop a two- Hybrid a computationally economical pra
stage multiple-kernel or ctical link artificial neural network
learning formula by Other (CEFLANN) and a rule set to a lot
incorporating consecutive token o AI of effectively
ptimization and therefore Techni generate trading choices.
the gradient projection technique. ques to They read the
Analyz stock mercantilism call as a
5 (201 Studies Chavan and Patil contribute to our e Stock classification downside with 3 pot
3) Using understanding of ANN stock Market ential values – get, hold or sell. The
Artifici exchange prediction s CEFLANN network employed
al by measure totally different model in the choice network produces a
Neural input parameters found collection of
Networ in 9 printed articles. They plan continuous mercantilism signals
ks to to realize the by analyzing
Predict foremost necessary input the nonlinear relationship that
Stock parameters exists between some well-
Market that manufacture higher model liked technical indicators. The
Values prediction output mercantilism signals are ac
) accuracy. supported their survey, customed track trends and to
they realize that almost provide mercantilism choices supp
all cc techniques build use of orted that
technical variables rather trend mistreatment mercantilism r
than elementary variables for a ules. this can be a
specific stock worth prediction, w unique approach centered on
hereas political profitable
economy variables are principally stock mercantilism choices throug
wont to predict exchange index h integration of the training ability
values. additionally, hybridized of the CEFLANN neural network
parameters manufacture higher res with the technical analysis rules.
ults in comparison with the 8 (201 Chong, Han and Park analyze
employment of solely one input 7) Studies deep learning networks for stock
variable kind. Using exchange analysis and prediction.
. Artifici Deep learning networks
6 (201 Studies The study by Patel, Shah, Thakkar al extract options from an
5) Using and Kotecha compares four Indian Neural oversized set of information while
Hybrid stock exchange prediction models: Networ not hoping
or ANN, SVM, random forest, and ks to on previous information of
Other naive-Bayes with 2 approaches for Predict predictors that makes it helpful for
AI model input. the primary approach Stock prime frequency stock
Techni for input files involves Market exchange prediction. they
ques to computation of 10 technical Values supply Associate in
Analyz parameters victimization stock co Nursing objective assessment
e Stock mmerce data (open, high, low and of each the benefits and
Market shut prices), whereas the second disadvantages of deep learning
s approach focuses on representing algorithms for stock
these technical parameters as exchange analysis and
trend settled knowledge. They prediction. exploitation high-
assess the accuracy of every of the frequency intraday stock returns
prediction models for every of the as computer file, they examine the
2 input approaches. analysis is results of 3 unattended feature
predicated on knowledge from 2 st extraction methods—
ocks and 2 stock value indices - principal element analysis,

IJCRTI020009 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 37

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882

autoencoder, and also

the restricted physicist machine—
on the network’s overall ability to
predict future market behavior.

III. PROPOSED METHODOLOGY

A. Yahoo Data
Stock market prediction seems a complex problem because
there are many factors that have yet to be addressed and it
doesn’t seem statistical at first. But by proper use of previous
data for machine learning techniques to make appropriate
assumptions and the current data and train the machine to
learn from it.
We used dataset for analysis was picked up from Yahoo
Finance. The dataset consisted of approximately 1 lakh
records of the required stock prices and other relevant values.
At certain time intervals for each day of the year the data
returns the stock prices. It made up of various sections
namely low, symbol, open, date, high and volume. The data
C. Long Short Term Memory(LSTM) Network Based
of only one company is considered for the purpose of
Model
simulation and analysis. First All the data is read which is
available in a file of csv format and transformed into a data- Long Short Term Memory is the advanced version of
frame using the keras, tensorflow in Python. From this, the Recurrent-Neural-Networks (RNN) where the information
data for one particular company is extracting by segregating belonging to previous state persists. These are different from
data on the basis of the symbol field. Following this RNNs as They involve long term dependencies and RNNs
normalization of the data was performed through usage of the works on finding the relationship between the recent and the
scikit library in Python and the data was divided into two current information so this are different from RNNs. it is
parts i.e testing and training sets. The test set was kept as 20% indicates that the interval of information is relatively smaller
of the available dataset. than that to LSTM.
Machine learning has many models but this paper focuses The predictions depends on large amounts of data and are
on four of the most important amongst them and made the generally dependent on the long term history of the market it
predictions using these. is the main purpose behind using LSTM model in stock
market prediction. LSTM regulates error by giving an aid to
B. Regression Based Models
In general, the Regression based Model is used for predicting
continuous values through some given independent values.
Regression uses a given linear function for predicting
continuous values:

V = a + bK + error [1]

Where, V is a continuous value; K represents known

independent values; and, a, b are coefficients. Fig. 1 Flow
Chart for Regression Based Mode.

The paper utilizes for predicting correct values by

minimizing the error function as given in Fig.1 the gradient
descent linear regression algorithm is uses. Linear Regression
algorithm as governed by equation [1] is performed on the
data and then the relevant predictions are made. Low, open,
high, close and volume are he factors considered for the
regression.

Fig 2 Flow Chart for LSTM

IJCRTI020009 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 38

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882

the RNNs through retaining information for older stages Firstly we import the required libraries like numpy
making the prediction more accurate Thus proving itself as ,pandas etc. then we load data and convert it into time series
much more reliable compared to other methods. data. Plotting of time series data is important to see the trend
Stock market involves processing of huge data, the and seasonality in data. After that splitting the data into
gradients with respect to the weight matrix may become very training and testing for evaluation purpose.
small and may degrade the learning rate. This corresponds to
the problem of Vanishing Gradient. LSTM prevents this from IV. CONCLUSION
happening. The LSTM consists of a remembering cell, input A. This paper provides a review on machine learning
gate, output gate and a forget gate. The cell remembers the techniques to predict share prices, In This review we
value for long term propagation and the gates regulate them. have surveyed an articles that have used artificial neural
network (ANN) and support vector machine (SVM) as a
D. Autoregressive Integrated Moving Average (ARIMA) forecasting model for stock prices. We also used LSTM
model algorithm, Linear Regression algorithm. LSTM
ARIMA model is short form of Autoregressive Integrated and Regression, on the Yahoo finance dataset. Both the
Moving Average. It is a class of models that 'explains' a techniques have shown an improvement in the accuracy
given time series based on its own past values, that is, its own of predictions, thereby yielding positive results. Use of
lags and the lagged forecast errors, so that equation can be recently introduced machine learning techniques in the
used to forecast future values. It converts non-stationary data prediction of stocks have yielded promising results and
into stationary data. thereby marked the use of them in profitable exchange
ARIMA is divided into three parts i.e. Auto schemes. It has led to the conclusion that it is possible to
Regression(AR), Integrated(I) and Moving Average(MA). predict stock market with more accuracy and efficiency
Auto Regression shows a regresses on its own lagged, or using machine learning techniques.
prior, values which is changing variables. Integrated shows
he differencing of raw observations to allow for the time
series to become stationary, i.e., data values are replaced by
the difference between the data values and the previous V. REFERENCES
values. Moving Average incorporates the dependency [1] National Stock Exchange
between an observation and a residual error from a moving https://www.nseindia.com/
average model applied to lagged observations. [2] Bombay Stock Exchange
https://www.bseindia.com/
[3] Stock market prediction using machine learning approach
https:// www.ijies.net/finial-docs/finial-pdf/220519CS11/
[4] API Keys
https:// www.alphavantage.co2
[5] Stock Market Prediction Using Artificial Neural
Networks
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.1
34.1114&rep=rep1&type=pdf/
[6] Stock market prediction using supervised machine
learning techniques
https://www.computer.org/csdl/proceedings-
article/csde/2020/09411609/1taFeSM9YBy/
[7] M. Usmani, S. H. Adil, K. Raza and S. S. A. Ali, "Stock
market prediction using machine learning techniques," 2016
3rd International Conference on Computer and Information
Sciences (ICCOINS)
https:// scialert.net/fulltext/?doi=jai.2018.48.54/
[8] "Regression techniques for the prediction of stock price
trend," 2012 International Conference on Statistics in
Science, Business and Engineering (ICSSBE)
https://ieeexplore.ieee.org/document/6396535/
[9] LSTM and Regression Methods for Stock Market
Prediction
ARIMA consists of three parameters which is nothing but p,
https://www.ijcrt.org/papers/IJCRT2005021.pdf
d and q. Where p stands for lag observation in model also
[10] Stock transaction prediction modeling and analysis
known as the lag model, d stands for the number of times that
based on LSTM
the raw observations are differenced; also known as the
https://www.researchgate.net/publication/326050300_Stock
degree of differencing and q stands for the size of the moving
_transaction_prediction_modeling_and_analysis_based_on_
average window; also known as the order of the moving
LSTM
average.
[11] Stock Price Prediction Using the ARIMA Model
https://ieeexplore.ieee.org/document/7046047

IJCRTI020009 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 39