Stock Market Prediction using Machine Learning

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 6 Issue 3, March-April 2022 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD49868 | Volume – 6 | Issue – 3 | Mar-Apr 2022 Page 2046
Stock Market Prediction using Machine Learning
Subham Kumar Gupta1
, Dr. Bhuvana J2
, Dr. M N Nachappa3
1
MCA Student, Department of Information Technology, JAIN Deemed-to-be University, Bengaluru, Karnataka, India
2
Head, Department of Information Technology, JAIN Deemed-to-be University, Bengaluru, Karnataka, India
3
Head, School of Computer Science & IT, JAIN Deemed-to-be University, Bengaluru, Karnataka, India
ABSTRACT
Stock market prediction is a typical task to forecast the upcoming
stock values. It is very difficult to forecast because of unbalanced
nature of stocks. In this work, an attempt is made for prediction of
stock market trend. This research aims to combine multiple existing
techniques into a much more robust prediction model which can
handle various scenarios in which investment can be beneficial. By
combing both techniques, this prediction model can provide more
accurate and flexible recommendations.
However instead of using those traditional methods, we approached
the problems using machine learning techniques. We tried to
revolutionize the way people address data processing problems in
stock market by predicting the behavior of the stocks. In fact, if we
can predict how the stock will behave in the short-term future we can
queue up our transactions earlier and be faster than everyone else. In
theory, this allows us to maximize our profit without having the need
to be physically located close to the data sources.
We examined three main models. Firstly we used a complete
prediction using a moving average. Secondly we used a LSTM model
and finally a model called ARIMA model. The only motive is to
increase the accuracy of predictive the stock market price. Each of
those models was applied on real stock market data and checked
whether it could return profit.
KEYWORDS: Stock market prediction
How to cite this paper: Subham Kumar
Gupta | Dr. Bhuvana J | Dr. M N
Nachappa "Stock Market Prediction
using Machine Learning" Published in
International Journal
of Trend in
Scientific Research
and Development
(ijtsrd), ISSN: 2456-
6470, Volume-6 |
Issue-3, April 2022,
pp.2046-2061, URL:
www.ijtsrd.com/papers/ijtsrd49868.pdf
Copyright © 2022 by author (s) and
International Journal of Trend in
Scientific Research and Development
Journal. This is an
Open Access article
distributed under the
terms of the Creative Commons
Attribution License (CC BY 4.0)
(http://creativecommons.org/licenses/by/4.0)
1. INTRODUCTION
Stock price prediction is very important as it is used
by most of the business people as well as common
people. People will either gain money or lose their
entire savings in stock market activity. Many Stock
holders invest more money on this stock market.
Without having a clear idea on stock market, many
people are losing a lot of money.
1.1. Motivation
Predicting the movement of stocks in the competitive
financial markets is a challenge even for the most
experienced day trader. Even in a fraction of a second
the price of a stock can change so drastically that the
first one who is able to see it and act can win huge
amount of money while the rest have to face a
financial disaster. Through the years many experts
used a variety of methods in order to try and predict
the unpredictable stock market and earn money.
1.2. Goals and Limitations
In this section we discuss the Goals and the
Limitations of this thesis. We explain in detail what
we want to achieve through the thesis and what
difficulties we had to overcome to make it happen.
1.3. Challenges and Limitations
Stock market is so complicated and many things can
affect the change in a price. Not only financial factors
can influence the price of a stock. Things like news or
the general mood can affect the price in many ways
positive or negative. If it was possible to model the
stock market with a function it would be a complex
function that lives in high-dimensional, maybe
infinite dimensional, space. Imagine what would
happen if someone knew a away to calculate that
function. That someone would be able to profit by
taking advantage of it. However the nature of the
IJTSRD49868

International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470
space is so complicated that finding that function is
an impossible thing to do. The real challenge is to try
and approximate that function using neural-networks
in a way that we can profit by applying it in the stock
market. The focus of this thesis is to try to
approximate the stock market as good as possible and
try to maximize our profit.
2. Literature Review
The stock returns is an area of study wherein many
research scholars have shown immense interest for
past several years. A brief review of literature will
help in understanding the relevance of the content
analysis in the area of stock returns. The first set of
articles includes studies that primarily focus on stock
market prediction using Moving Average, ARIMA
MODEL and LSTM. The researches in social
sciences or in the field of economics depend in one
way or the other on careful reading of written
materials and the research work done by many
research scholars on similar subjects. Considering this
fact, the importance of content analysis becomes very
significant.
The objective of this paper is to construct a model to
predict stock worth movement mistreatment the
Moving Average, ARIMA MODEL and LSTM to
predict National securities market (NSE). It used
domain specific approach to predict the stocks from
every domain and brought some stock with most
capitalization. Topics and connected opinion of
shareholders are mechanically extracted from the
writings in an exceedingly message board by utilizing
our projected strategy aboard uninflected clusters of
comparable type of stocks from others mistreatment
clump algorithms.
The various areas to which the technique of content
analysis can be applied are based on the user’s skill
and ingenuity in framing valid category formats as
discussed in the research. Stock price prediction is a
challenging task owing to the complexity patterns
behind time series. Autoregressive integrated moving
average (ARIMA) model, Moving Average and
Long-Short Term Memory (LSTM) model are
popular linear and nonlinear models for time series
forecasting respectively. The integration of two
models can effectively capture the linear and
nonlinear patterns hidden in a time series and improve
forecast accuracy. In this paper, a new hybrid
ARIMA-BPNN model containing technical indicators
is proposed to forecast four individual stocks
consisting of both main board market and growth
enterprise market in software and information
services.
Barelson (1952) defined content analysis as a
technique of research that is systematic representation
of the matter of communication. According to Stone
(1964), the content analysis is a methodology or
procedure which can be used to access particular
information based on the past references. The
definition of content analysis requires that the
inference be derived from the counts of frequency to
place a number of standard methods on the borderline
of acceptability (Leites & Poo, 1942).
Enke and Thawornwong (2005) use a machine
learning information gain technique to evaluate the
predictive relationships for numerous financial and
economic variables. By computing the information
gain for each model variable, a ranking of the
variables is obtained. A threshold is determined to
select only the strongest relevant variables to be
retained in the forecasting models.
2.1. SCOPE OF THE STUDY
Data forecasting is really convenient topic of research
from last few decades and may remain active topic in
upcoming years also. Stock price prediction also used
data forecasting is basically is one of favorite topic
among researchers. A lot of research is already done
in this field like Stock price prediction can be using
neural, fuzzy, machine learning, R programming and
so on. Here we want to predict future stock price
using the some predictive services. This will provide
help to get more accurate results for predicting stock
price prediction. In future we can analyze this stock
market historical data in some other way to find more
accurate results. We can deal with the not only
finding of future stock price prediction but also tried
to reduce mismatch value i.e. difference between
actual price and predicted price. Threshold value can
be reduced to move toward more accurate value.

3. Prediction Techniques
Fig 1: Prediction Techniques
3.1. MOVING AVERAGE
Fig 2: Moving Average
‘Average’ is easily one of the most common things we use in our day-to-day lives. For instance, calculating the
average marks to determine overall performance, or finding the average temperature of the past few days to get
an idea about today’s temperature – these all are routine tasks we do on a regular basis. So this is a good starting
point to use on our dataset for making predictions.
3.2. ARIMA MODEL
Auto-Regressive Integrated Moving Average is a model which is used in statistics and econometrics to measure
events that happen over a period of time. It’s a class of statistical models for analyzing and forecasting time
series data. The model understands past data and predicts future data in the series. It’s used when a metric is
recorded in regular intervals, from fractions of a second to daily, weekly or monthly periods.
AR – Auto-regression: It predicts future values based on past values.
I - integrated: The use of differencing of raw observations in order to make the time series stationary
MA - Moving Average: It is the dependency between an observed value and a residual error from a moving
average model applied to previous observations.
In forecasting stock prices, the model reflects the differences between the values in a series rather than
measuring the actual values.
ADVANTAGES:
∑ ARIMA model has a fixed structure and is specifically built for time series (sequential) data.
∑ ARIMA works better for relatively short series when the number of observations is not sufficient to
apply more flexible methods.
DISADVANTAGES:
∑ ARIMA models can only be highly accurate and reliable under the appropriate conditions and data
availability.
∑ The ARIMA model tends to be unstable, both with respect to changes in observations and changes in
model specification.

3.3. LSTM – Long Short – Term Memory
Long Short-Term Memory (LSTM) networks are a type of recurrent neural network capable of learning order
dependence in sequence prediction problems as they can store past information. This is important as the previous
price of a stock is crucial in predicting its future price. LSTM neural networks are capable of solving numerous
tasks that are not solvable by previous learning algorithms like RNNs. Long Short-Term Memory (LSTM)
networks are a modified version of recurrent neural networks, which makes it easier to remember past data in
memory. LSTM is well-suited to classify, process, and predict time series given time lags of unknown duration.
A common LSTM unit is composed of a cell, an input gate, an output gate and a forget gate.
Fig 3: LSTM
ADVANTAGES:
∑ LSTM cells have a memory that can store previous time step information and use it to train the dataset. It
has the ability to bridge very long time lags.
∑ LSTM doesn’t have the vanishing gradient problem which a traditional RNN has.
DISADVANTAGES:
∑ They require a lot of resources and time to get trained and become ready for real-world applications
What is GCP based on?
What is GCP? GCP is a public cloud vendor — like competitors Amazon Web Services (AWS) and Microsoft
Azure. With GCP and other cloud vendors, customers are able to access computer resources housed in Google's
data centers around the world for free or on a pay-per-use basis.
4. Financial Definition
4.1. Stock Market
Stock prediction is using historical price, related market information and so on to forecast exact price or price
trend of the stock in the near future. According to the time granularity of price information, stock trading can be
divided into low-latency trading based on daily basis and high-frequency trading, which market exchanges in a
matter of hours, minutes, even seconds. High-frequency trading analysis is more common in hedge funds,
investment banks, and large institutional investors. It masters the trading signals before prices' ups and downs
through analyzing great amount of trading data [20]. In this thesis, only low-latency trading is taking into
consideration, which is more common in academia. Its core concept is to increase the accuracy of stock
prediction based on the related market information
The Indian stock market mainly studies stocks traded at National Stock Exchange (NSE) and Bombay Stock
Exchange (BSE). NSE or National Stock Exchange is located in Mumbai, and it is India’s leading stock
exchange market. It first came into existence in 1992 and brought with it an electronic exchange system in India,

which led to the removal of the paper based system. In 1875, BSE or Bombay Stock Exchange was established,
and it was formerly known as 'The native share and stock brokers association’. However, after 1957,
Government of India recognized this stock exchange as the premier stock exchange of India, under the Securities
Contract Regulation Act, 1956.
The BSE's Sensex comprises of 30 companies, while NSE's Nifty comprises of 50 companies.
Both the stock exchanges, National Stock Exchange and Bombay Stock Exchange, are an important part of
Indian Capital Market. Every day, hundreds of thousands of brokers and investors trade on these stock
exchanges. And both are established in Mumbai, Maharashtra, and SEBI (Securities and Exchange Board of
India) recognized.
4.2. Stock Trend Definition
In this thesis, I will mainly focus on predicting ups and downs of stock. Stocks leave some important trading
data after each trading day, such as open price, close price, adjusted close price, highest price, lowest price,
volume, etc. Among all, adjusted close price usually represents the stock price of that trading day. In the trading
period, there will be a series of adjusted close prices. Let us denote it as:
p1, p2, p3,..., pT .
Here, pt is the close price on t trading day, T is total trading days in this period. Stock price of a certain trading
day will rise or drop comparing to previous trading day, thus, here I used the change of closing price of two
consecutive trading days as the judgment. Let us denote trading situation as:
yt = 8 ><>: 1 if pt > pt1, 0 if pt pt1.
If it is 1, it means the price goes up on the second trading day. Otherwise, if it is 0, it means the price goes down
or remains the same.
5. Objective of the Study
Objectives of the study are defined as follow:
The main objective of this study is to predict the future stock price by analyzing the past historical data that
we were going to collect from the National Stock Exchange.
Predicting the Stock market cost in such a way that it will provide most accurate results.
Stock market price forecasting should be done in such a way that predicted price should minimize the
threshold value (difference between actual value and predicted value also known as mispricing) and close
enough to the actual value.
Process of analyzing the historical data should be simple and easy to understand. For this feature
identification can done intelligently to provide most accurate results.
To increase the efficiency of the data analysis technique by using some cloud based tools.
To analyze the performance and comparing proposed algorithm with the existing algorithms in terms of
predicted price accuracy, close price predicted and accurate close price etc.
6. Methodology
6.1. PROBLEM STATEMENTS
Stock market is so complicated and many things can affect the change in a price. Not only financial factors can
influence the price of a stock. Things like news or the general mood can affect the price in many ways positive
or negative. If it was possible to model the stock market with a function it would be a complex function that
lives in high-dimensional, maybe infinite dimensional, space.
Imagine what would happen if someone knew a away to calculate that function. That someone would be able to
profit by taking advantage of it. However the nature of the space is so complicated that finding that function is
an impossible thing to do. The real challenge is to try and approximate that function using neural-networks in a
way that we can profit by applying it in the stock market. The focus of this thesis is to try to approximate the
stock market as good as possible and try to maximize our profit.
We’ll dive into the implementation part of this article soon, but first it’s important to establish what we’re
aiming to solve. Broadly, stock market analysis is divided into two parts – Fundamental Analysis and Technical
Analysis.
Fundamental Analysis involves analyzing the company’s future profitability on the basis of its current
business environment and financial performance.

Technical Analysis, on the other hand, includes reading the charts and using statistical figures to identify the
trends in the stock market.
As you might have guessed, our focus will be on the technical analysis part. We’ll be using a dataset from data-
flair. tranings (you can find historical data for various stocks here) and for this, I have used the data for ‘Tata
Global Beverages’.
Another difficulty we had to face was the way to determine winnings. We had two different options.
1. Winnings is the difference between portfolio value plus the capital we have on our possession and the initial
capital.
2. Winnings is the sum of all the differences in price between sequential transactions.
For example if we bought a stock at price x and sold it at price y, then the winnings are y-x.
Fig 4: Data Flow
6.2. IMPLEMENTATION
We will first load the dataset and define the target variable for the problem.
Datasets - We will implement this technique on our dataset. The first step is to create a data-frame that contains
only the Date and Close price columns, then split it into train and validation sets to verify our predictions.
Fig 5: Tata Global Dataset

There are multiple variables in the dataset – date, open, high, low, last, close, total_trade_quantity, and turnover.
The columns Open and Close represent the starting and final price at which the stock is traded on a particular
day.
High, Low and Last represent the maximum, minimum, and last price of the share for the day.
Total Trade Quantity is the number of shares bought or sold in the day and Turnover (Lacs) is the turnover
of the particular company on a given date.
Another important thing to note is that the market is closed on weekends and public holidays. Notice the above
table again, some date values are missing – 2/10/2018, 6/10/2018, 7/10/2018. Of these dates, 2nd is a national
holiday while 6th and 7th fall on a weekend. The profit or loss calculation is usually determined by the closing
price of a stock for the day; hence we will consider the closing price as the target variable.
6.2.1. MOVING AVERAGE:
In stock market analysis, a 50 or 200-day moving average is most commonly used to see trends in the stock
market and indicate where stocks are headed. The MA is used in trading as a simple technical analysis tool that
helps determine price data by customizing average price. There are many advantages in using a moving average
in trading that can be tailored to any time frame. Depending on what information you want to find out, there are
different types of moving averages to use.
The MA is the calculated average of any subset of numbers, using a technique to get an overall idea of the trends
in a data set. Once you understand the MA formula, you can start to calculate any subsets to get your MA. It can
be calculated for any period of time, making it extremely useful to forecast both long and short-term trends.
Fig 6: Moving Average (SMA 20)
The SMA formula is calculated by taking the average closing price of a security over any period desired. To
calculate a moving average formula, the total closing price is divided by the number of periods.
For example, if the last five closing prices are:
28.93+28.48 +28.44+28.91+28.48 = 143.24
The five-day SMA is: 142.24/5= 28.65.
6.2.2. ARIMA MODEL:
We can de-trend the model by differencing each value from a value in the past and modeling these differences.
Later, adding the value from the past to arrive at the actual value. We can choose to difference each value from a
value at t-1 or t-2 or t-3 … Upon experimentation I found t-2 to give good results. By good results I mean the
stationarity of the resulting time series was better. Again, to know what is stationarity and how to measure it
please go through the link in the prerequisite section.

Fig 7: Differenced Close Price
Auto Regress or (p) Integrated (d) Moving Average(q).
p — Number of previous values to consider for estimating the current value
d — n_diff in the previous code snippet
q — If we consider a moving average to estimate each value, then q indicates the number of previous errors.
i.e., if q= 3 then we will consider e(t-3), e(t-2) and e(t-1) as inputs of the regressor.
Where e(i) = moving_average(i)- actual_value(i)
6.2.3. LONG-SHORT TERM MEMORY (LSTM):
Fig 8: LSTM Cells
The first step in our LSTM is to decide the information to throw away from the cell state. The decision is made by
a sigmoid layer called the “forget gate layer.”
The next step is to decide what new information is going to store in the cell state. It has two parts. First, a sigmoid
layer called the “input gate layer”. It decides which values will be updated. Next, a tanh layer creates a vector of
new candidate values, C~t C~t, that could be added to the state.
Finally, output is decided and it’ll be based on our cell state, but will be a filtered version.
7. Coding
7.1. Stock Price prediction using Moving Average
#import dataset
from google.colab import files
uploaded = files.upload()

#import packages
import pandas as pd
import numpy as np
#to plot within notebook
import matplotlib.pyplot as plt
%matplotlib inline
#setting figure size
from matplotlib.pylab import rcParams
rcParams['figure.figsize'] = 20,10
#for normalizing data
from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler(feature_range=(0, 1))
#read the file
df = pd.read_csv('NSE-Tata-Global-Beverages-Limited.csv')
#print the head
df.head()
#setting index as date
df['Date'] = pd.to_datetime(df.Date,format='%Y-%m-%d')
df.index = df['Date']
#plot
plt.figure(figsize=(16,8))
plt.plot(df['Close'], label='Close Price history')
# importing libraries
import pandas as pd
import numpy as np
# reading the data
# looking at the first five rows of the data
print(df.head())
print('n Shape of the data:')
print(df.shape)
# setting the index as date
#creating dataframe with date and the target variable
data = df.sort_index(ascending=True, axis=0)
new_data = pd.DataFrame(index=range(0,len(df)),columns=['Date', 'Close'])
for i in range(0,len(data)):
new_data['Date'][i] = data['Date'][i]
new_data['Close'][i] = data['Close'][i]
# NOTE: While splitting the data into train and validation set, we cannot use random splitting since t
hat will destroy the time component. So here we have set the last year’s data into validation and the 4
years’ data before that into train set.
# splitting into train and validation
train = new_data[:987]
valid = new_data[987:]
# shapes of training set
print('n Shape of training set:')
print(train.shape)

# shapes of validation set
print('n Shape of validation set:')
print(valid.shape)
# In the next step, we will create predictions for the validation set and check the RMSE using the act
ual values.
# making predictions
preds = []
for i in range(0,valid.shape[0]):
a = train['Close'][len(train)-248+i:].sum() + sum(preds)
b = a/248
preds.append(b)
# checking the results (RMSE value)
rms=np.sqrt(np.mean(np.power((np.array(valid['Close'])-preds),2)))
print('n RMSE value on validation set:')
print(rms)
#plot the graph
valid['Predictions'] = 0
valid['Predictions'] = preds
plt.plot(train['Close'])
plt.plot(valid[['Close', 'Predictions']])
7.2. Stock Price prediction using LSTM :
#import dataset
#import packages
import pandas as pd
import numpy as np
%matplotlib inline
#read the file
#print the head
df.head()
#plot
#importing required libraries
from keras.models import Sequential
from keras.layers import Dense, Dropout, LSTM

#creating dataframe
new_data = pd.DataFrame(index=range(0,len(df)),columns=['Date', 'Close'])
for i in range(0,len(data)):
new_data['Date'][i] = data['Date'][i]
new_data['Close'][i] = data['Close'][i]
#setting index
new_data.index = new_data.Date
new_data.drop('Date', axis=1, inplace=True)
#creating train and test sets
dataset = new_data.values
train = dataset[0:987,:]
valid = dataset[987:,:]
#converting dataset into x_train and y_train
scaled_data = scaler.fit_transform(dataset)
x_train, y_train = [], []
for i in range(60,len(train)):
x_train.append(scaled_data[i-60:i,0])
y_train.append(scaled_data[i,0])
x_train, y_train = np.array(x_train), np.array(y_train)
x_train = np.reshape(x_train, (x_train.shape[0],x_train.shape[1],1))
# create and fit the LSTM network
model = Sequential()
model.add(LSTM(units=50, return_sequences=True, input_shape=(x_train.shape[1],1)))
model.add(LSTM(units=50))
model.add(Dense(1))
model.compile(loss='mean_squared_error', optimizer='adam')
model.fit(x_train, y_train, epochs=1, batch_size=1, verbose=2)
#predicting 246 values, using past 60 from the train data
inputs = new_data[len(new_data) - len(valid) - 60:].values
inputs = inputs.reshape(-1,1)
inputs = scaler.transform(inputs)
X_test = []
for i in range(60,inputs.shape[0]):
X_test.append(inputs[i-60:i,0])
X_test = np.array(X_test)
X_test = np.reshape(X_test, (X_test.shape[0],X_test.shape[1],1))
closing_price = model.predict(X_test)
closing_price = scaler.inverse_transform(closing_price)
#calculating RMS value
rms=np.sqrt(np.mean(np.power((valid-closing_price),2)))
rms
#for plotting
train = new_data[:987]
valid = new_data[987:]
valid['Predictions'] = closing_price
plt.plot(valid[['Close','Predictions']])

7.3. Stock Price prediction using ARIMA Model:
#import dataset
#import packages
import pandas as pd
import numpy as np
%matplotlib inline
#read the file
#print the head
df.head()
#plot
#installing new lib files
pip install pyramid
pip install pmdarima
#import package and setting the training
from pmdarima import auto_arima
train = data[:987]
valid = data[987:]
training = train['Close']
validation = valid['Close']
model = auto_arima(training, start_p=1, start_q=1,max_p=3, max_q=3, m=12,start_P=0, seasonal=T
rue,d=1, D=1, trace=True,error_action='ignore',suppress_warnings=True)
model.fit(training)
forecast = model.predict(n_periods=248)
forecast = pd.DataFrame(forecast,index = valid.index,columns=['Prediction'])
#calculating RMS value
rms=np.sqrt(np.mean(np.power((np.array(valid['Close'])-np.array(forecast['Prediction'])),2)))
rms
#plot the graph
plt.plot(valid['Close'])
plt.plot(forecast['Prediction'])

8. Results
8.1. Moving Average Predicted Price
Let’s visualize this to get a more intuitive understanding. So here is a plot of the predicted values along with the
actual values.
Fig 9: Close price (MA)
Fig 10: Predicted Price (MA)
8.2. ARIMA Model Predicted Price
ARIMA model uses past data to understand the pattern in the time series. Using these values, the model captured
an increasing trend in the series. Although the predictions using this technique are far better than that of the
previously implemented machine learning models. As it’s evident from the plot, the model has captured a trend
in the series, but does not focus on the seasonal part.
Fig 11: Close Price (ARIMA)

Fig 12: Predicted Price (ARIMA)
8.3. LSTM (Long-Short Term Memory) Predicted Price
The LSTM model can be tuned for various parameters such as changing the number of LSTM layers, adding
dropout value or increasing the number of epochs. At the start of the article, stock price is affected by the news
about the company and other factors like demonetization or merger/demerger of the companies. There are certain
intangible factors as well which can often be impossible to predict beforehand.
Fig 13: Close Price (LSTM)
Fig 14: Predicted Price (LSTM)

8.4. Comparative Analysis of Machine Learning Techniques:
S.
No.
Techniques Advantages Disadvantages
Parameter
Used
1
Moving
Average
The SMA is the most
straightforward calculation, the
average price over a chosen time
period.
The SMA’s weakness is that it
is slower to respond to rapid
price changes that often occur
at market reversal points.
Close
Price of
stocks.
2
ARIMA
MODEL
ARIMA works better for relatively
short series when the number of
observations is not sufficient to
apply more flexible methods.
It is suitable for short-term
prediction only.
Open,
High,
Close,
Low prices
and
moving
average.
3
LSTM(Long-
Short Term
Memory)
LSTM cells have a memory that can
store previous time step information
and use it to train the dataset. It has
the ability to bridge very long time
lags.
They require a lot of resources
and time to get trained and
become ready for real-world
applications.
Open,
High,
Close,
Low
prices.
8.5. RMSE : Root-Mean-Square Error
RMSE is defined as the square root of the average squared distance between the actual score and the predicted
RMSE is used to evaluate the Machine Learning models. So here,
Moving average – 104.51415465984348
ARIMA model – 44.954584993246954
LSTM – 9.707045241044804
9. Conclusion
This work summarizes necessary techniques in
machine learning that square measure relevant to
stock prediction. This model may be improved upon
by process refined fuzzy rules. By coaching data’s
scale and timeframe may result in higher prediction.
Technical indicators square measure accustomed
construct the relation between stock exchange index
and their variables. Implementing victimization
Moving Average slow so as to perform computations
compared to alternative techniques, whereas ARIMA
model is good for short-run prediction. Every
technique has its own benefits and downsides.
Differing types of techniques are accustomed predict
the stock exchange and to forecast the longer term
stock values up to some extent. Combining Moving
Average, ARIMA Model and LSTM Model might
end in high accuracy.
9.1. General Comment
We were able to come with a way to successfully
predict the stock market and in combination with a
good trading strategy we were able to profit from
stock trading using historical data. The reason we
used historical data and not real time data for testing
was time efficiency but also the ability to compare
models and trading strategies using the same testing
data.
How would the model behave with real-time data?
We treated our historical data as real time data. We
can use the same methods and be able to predict the
stock price in real time. We can achieve this by
collecting the transactions in real time and converting
them into 5-min intervals and just passing them
forward to our network, point by point. One of the
goals of this thesis was that we should be able to
utilize the stock market on real time and all the
simulations were done in way that would make it easy
to transition from historical to real time data.
As an investment it can be characterized as really
profitable. However the neural network cannot
predict sudden changes in the price that happen
during the time that the stock market is closed. An
example is when a company or their direct
competitors announce their term results. Those kinds
of events can skyrocket the stock price or make it lose
considerable value.

9.2. Limitations
The main limitation we had during this thesis was the
time constraints due to the lack of available data. We
had to wait at least three months to run experiments
as we had to collect our own data. Also trying to train
a neural network takes quite long. If we were lucky a
neural network would converge in a couple of hours
but sometimes could take up to 15 hours. Especially
in the Reinforcement Learning (RL) models training
would take quite a few days. The whole process it
was a trial and error in order to come up with the
optimal hyper parameters and we had to train numeral
different networks just to be able to select them.
Stock market cannot be accurately predicted. The
future, like any complex problem, has far too many
variables to be predicted. The stock market is a place
where buyers and sellers converge. When there are
more buyers than sellers, the price increases
9.3. Future Work
We have to test the existing methods with more data
as the keep coming. We want to ensure that the
results we got it is not just a random event that
happened as result of the time period. We have to test
with even more data as time passes and make sure
that our model can generalize. Another thing we can
do is check how our model works with stocks outside
the OMSX30 index. We can try train and evaluate out
model with stocks that belong to smaller companies
that do not have as many transactions as the big ones.
In that case we will be able to see if we can expand
our work to other stocks and maybe even other stock
markets as well.
The concept behind this idea is that we will have few
neural networks trained in different time intervals.
For example, we can have the prediction of the stock
price in 5-min, 10-min and 30-min intervals. Using
this information we can decide when is the best time
to place our action. If we know how the stock will
move within the next thirty minutes we will be able to
increase our profit even more. The more information,
we have the more profit we can achieve.
REFERENCES
[1] Prashant S. Chavan,”Parameters for Stock
Market Prediction” IJCTA, MAR 2013.
[2] Mayank Kumar B Patel, Sunil R Yalamalle
“Stock price prediction using Artificial Neural
Network “IJIRSET, June 2014.
[3] Mrityunjay Sharma “Survey on Stock Market
Prediction and Performance Analysis”
IJARCET, January 2014.
[4] Gholap Rahul Mansing “Indian stock market
prediction using neural network technique”
IJARCET, March 2014.
[5] Shom Prasad Das, Sudarsan Padhy “Support
Vector Machines for Prediction of Futures
Prices in Indian Stock Market” IJCA, March
2014.
[6] Author: Yusuf Perwej, Asif Perwej Research
paper: Prediction of the Bombay Stock
Exchange (BSE) Market Returns Using
Artificial Neural Network and Genetic
Algorithm Journal: Scientific Research
[7] Author: K. Senthamarai Kannan, P. Sailapathi
Sekar, M. Mohamed Sathik and P. Arumugam
Research paper: Financial Stock Market
Forecast using Data Mining Techniques
Journal: International Multi-Conference of
Engineers and Computer Scientists 2010 Vol I,
IMECS 2010, March 17-19, 2010, Hong Kong.
ISSN:2078-0966
[8] Author: Zahid Iqbal, R. Ilyas, W. Shahzad, Z.
Mahmood and J Anjum Research paper:
Efficient Machine Learning Techniques for
Stock Market Prediction Journal: Int. Journal of
Engineering Research and Applications, ISSN:
2248-9622, Vol. 3, Issue 6, Nov-Dec 2013,
pp.855-867.
[9] Author: Marc-André Mittermaye Research
paper: Forecasting Intraday Stock Price Trends
with Text Mining Techniques Journal: Hawaii
International Conference on System Sciences –
2004.
[10] Hellstrom, T. and Holmstrom, K. 1998,
'Predicting the Stock Market', Technical Report
Series Ima-TOM-1997-07.
[11] 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), Kuala Lumpur, 2016, pp.
322-327.

Stock Market Prediction using Machine Learning

More Related Content

What's hot (20)

Similar to Stock Market Prediction using Machine Learning (20)

More from ijtsrd (20)

Recently uploaded (20)

Stock Market Prediction using Machine Learning