See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/341322930

Stock Price Trend Prediction Using Multiple Linear Regression

Article · May 2020

CITATIONS READS

10 8,150

5 authors, including:

Pratik Kanani
Dwarkadas J. Sanghvi College of Engineering
89 PUBLICATIONS 342 CITATIONS

SEE PROFILE

All content following this page was uploaded by Pratik Kanani on 12 May 2020.

The user has requested enhancement of the downloaded file.

International Journal of Engineering Science Invention (IJESI)
ISSN (Online): 2319 – 6734, ISSN (Print): 2319 – 6726
www.ijesi.org ||Volume 7 Issue 10 Ver II || Oct 2018 || PP 29-33

Stock Price Trend Prediction Using Multiple Linear Regression

Shruti Shakhla1, Bhavya Shah1, Niket Shah1, Vyom Unadkat1
Pratik Kanani2
1
(Student, Information Technology Engineering,
Dwarkadas J. Sanghvi College of Engineering, Mumbai University, India)
2
(Assistant Professor, Information Technology Engineering,
Dwarkadas J. Sanghvi College of Engineering, Mumbai University, India)
Corresponding Author: Shruti Shakhla

Abstract : Stock markets are major institutions where trades are made daily having their worth in millions and
billions on a daily basis. Some consider as a fast way to fill up their pockets whereas some have a traditional
approach to invest in a company and reap long term benefits. A well done research about the company and its
performance can help to gain better financial profits. Industry experts and major conglomerates invest heavily
in research and development and with the technological developments, minimising the error probability and
estimating the future performance has become achievable. Machine Learning algorithms have proved to fetch
benevolent results in predicting stock prices. In this paper, we have studied and documented the performance of
APPLE INC.’s stock price using Multiple Linear Regression and gauged its performance using Root Mean
Squared Error. The results are promising but can be improved by taking into consideration more parameters.
Keywords -Stock Market, Finance, Linear Regression, Machine Learning
----------------------------------------------------------------------------------------------------------------------------- ----------
Date of Submission: 05-10-2018 Date of acceptance: 17-10-2018
----------------------------------------------------------------------------------------------------------------------------- ----------

I. INTRODUCTION
Making money hand over fist is the real incentive which draws investors towards the stock market. The
stock market is a platform where individuals can buy and sell shares of publicly traded companies. It is a system
with individuals of opposite views, ideas and approach. One investor will sell a particular stock whereas
someone else will buy it. One of them would profit from the trade and hence, it is very important to be well
versed while dealing with stocks to maintain a profitable account.[1]
There are many factors affecting the price of the stock be it supply and demand, media, risk, statements
made by the market experts, political moves, natural disasters and many more. All of these and an array of other
factors together when compiled, have a direct impact on the stock price. Investors need to study extensively
before investing in a particular stock. The stock’s actual price and its fair value need to be taken into
consideration to make crucial decisions. The actual price is the price at which it is traded in the market. The fair
value can be estimated by the company’s assets keeping the liabilities aside. It can also be derived by the net
value of the company’s future earnings.
The indispensable surge towards prediction models has compelled financial analysts in adopting
machine learning models and computational intelligence. Technical analysis of stock markets is made easier by
implementing machine learning models such as Regression, Support Vector Regression and Deep Networks
with the help of various tools and libraries.[2]
Here we use a linear regression model to learn the trends of the market and then we predict the price of
the stock, to givea brief idea about how the stock is going to perform.

II. RELATED WORK

R. Yamini and Dr. C. Dhaya highlighted the major factors that are to be considered while selecting the
best suited Machine Learning Prediction Model for stock analysis. They used multiple prediction models to
compare and analyze the performance of each model in order to find the most accurate model that must be
implemented for stock analysis.[3]
Another paper written by Han Lok Siew and Md Jan Nordin demonstrated the importance of using
ordinal or standardized values from a dataset made of heterogeneous types of data, in regression prediction
techniques. The paper hypothesized the factors to be considered for a prediction model, which conjugated the
basis of performing fundamental analysis of stocks. The variables were extracted from balance sheets, income
statements and other cash flow statements provided in the annual reports of the particular stock tickers.[4]

www.ijesi.org 29 | Page
 Stock Price Trend prediction using Multiple Linear Regression

K.R. Sekar, Manav Srinivasan, K.S.Ravichandran and J.Sethuram wrote a paper on the prediction of
gold price in the stock market based on several independent yet influential variables. The prediction model used
multiple linear regression algorithm to predict the price of gold in the market. Their model took a dataset
consisting of historical gold prices, along with other variables of many years on a monthly basis to feed into
their model which would be used for prediction later on.[5]

III. PROBLEM DEFINITION

Nature of the stock market is extremely volatile. The major goal is to minimize the uncertainty of the
returns by meticulously predicting (as accurately as possible) the future stock prices and thereby identifying
their fluctuations in advance to mitigate financial risks. For the following reasons we have developed, our own
prediction model which takes into account multiple factors in order to predict the stock prices.

IV. DATA METHODOLOGY

Figure 1: Methodology

The stocks considered for our prediction model are AAPL(Apple), which comes under the market
index NASDAQ(NDAQ). Our prediction model will be trained on the historical data gathered using the Yahoo
Finance Developer API.[6][7][8]
The Yahoo API provides various methods which facilitates data acquisition. Using this API, we focus
on gathering the following two data parameters:
 Stock’s Day Open Price
 Stock’s Day High Price; for both, AAPL and NDAQ.

Further, since the stock prices do not change over the weekend, we wish to eliminate them. Yahoo API
provides a feature which helps acquire only weekday’s stock prices. After acquiring the requisite data, it is
significant to follow it up with data-preprocessing. To aid this process, the data gathered is converted to Panda’s
Data-frame and indexed. Furthermore, it is noticed that there are various NAN (Not A Number) values in the
Data-frame. A NAN value in a Panda’s Data-frame is indicative of a missing value. These missing values make
www.ijesi.org 30 | Page
 Stock Price Trend prediction using Multiple Linear Regression

the Data-frame inconsistent and prevent application of any predictive models on them. There are various ways
to deal with these NAN values and we chose to use the panda’s function fillna() on the Data-frame to forward-
fill and backward-fill the values. Forward-fill uses last valid observation to fill missing values up to the next
valid observation whereas backward-fill uses the next valid observation to fill in the gap. After eliminating all
redundancies and inconsistencies in the data, the predictive model can now be trained using it.[9]

V. PROPOSED MODEL
Multiple linear Regression [10] is a highly established statistical technique used in stock market
analysis. It allows the analyser to consider multiple variables which affect the quantity to be predicted. The
quantity to be predicted is usually referred to as the independent variable. The various factors which
demonstrate a strong correlation with the independent variable are referred to as the dependent variables. The
measure by which the dependent variable changes due to a unit change in the independent variable is known as
the regression coefficient of that independent variable.
This model mathematically calculates a linear (straight line) relation between the dependent variable
and every other independent variable (Apple’s daily high and Market’s daily high). The general mathematical
equation representing a multiple linear regression is:
Yi = B0 + B1Xi1 + B2Xi2 + … + BnXin + E[11]
where Yi : dependent variable
Xi1, Xi2, …, Xin : n-independent variables
B1, B2, …, Bn : regression coefficients
B0 : y-intercept
E : random error in the
model
One of the popular use of this model is considering Least Squares Regression. Least Squares
Regression aims at calculating a best fit line by minimising the residual sum of squares of the deviations of the
predicted values from the corresponding data points. Squaring the deviations removes the complexity introduced
in the model due to positive and negative values. Based on above information we have formulated our own
prediction model to carefully predict a stock’s daily high price.

Figure 2: Stock Prediction Model

The Prediction Model using Multiple Linear Regression Method has been built using Python
Programming. We aim to predict a stock’s daily high using historical data. The data used is the stock’s open and
the market’s open. The model used is a Multi-Linear Regression model which is one of the most extensively
used model in stock market analysis.
Considering the dynamic nature of the stock market, the opening price of a stock on a particular day is
highly dependent on a plethora of factors. Hence considering the opening price of a stock as an independent
variable for our prediction model will give us an amalgamation of all those factors. The market’s opening price
gives us a broader understanding of the trends of the stock, especially in case of a drastic fluctuation in the entire
market.
The stock considered is AAPL, which comes under the market index Nasdaq (NDAQ). Historical data
is gathered using Yahoo Finance [6].

www.ijesi.org 31 | Page
 Stock Price Trend prediction using Multiple Linear Regression

VI. FINDINGS AND DISCUSSIONS

The model has been trained on data of 12 months.
 Training Dates: 30th September, 2016 – 30th September, 2017
 Test Dates: 1st October, 2017 – 31st October, 2017
Table 1: Table of regression coefficients and intercept
AAPL Open NDAQ Open
Intercept
Coefficient Coefficient

˗0.25142712813 0.99658308 0.02304602

APPLHigh = -0.251427128131 + (0.99658308 * AAPLOpen) + (0.02304602 * NDAQOpen)

Figure 3: Predicted vs Observed values

This model produces a R2 score of 0.91433103366658208 and Root Mean Square Error (RMSE) of
1.145040040250809. The R2 score, also known as the coefficient of determination is the proportion of variance
in the dependent variable that can be predicted from the independent variables. The Root Mean Square Error
(RMSE) is a frequently used measure of the differences between values predicted by a model or an estimator
and the values actually observed.

VII. CONCLUSION
Modelling of capital markets has authentically been done in partial equilibrium. Such machine
learning models indeed provide us with a broader understanding of the capital market. Hence we have harnessed
this opportunity to gain an insight about the market. In this paper we have incorporated the multiple linear
regression technique, which is one of the most routinely used technique to make predictions in the capital
market. Considering the open price of the stock as an independent variable eliminates the need to consider other
factors to predict the high of the stock. The open of the market provides with an insight of the variations in stock
prices of the entire market. Taking in account the above mentioned factors, the linear regression model produced
a RMSE of 1.145040040250809.
We believe that the performance measures obtained here can be improved further with the use of
genetic-based algorithms [12-18]. Taking into consideration a wide array of factors could widen the scope of the
model.

REFERENCES
[1]. The Economic Times, “Definition of Stock Market” https://economictimes.indiatimes.com/definition/stock-market
[2]. Sebastien Heiz, “A simple deep learning model for stock price prediction using TensorFlow” https://medium.com/mlreview/a-
simple-deep-learning-model-for-stock-price-prediction-using-tensorflow-30505541d877
[3]. R. Yamini Nivetha, Dr. C. Dhaya, “Developing a Prediction Model for Stock Analysis” 2017 International Conference on Technical
Advancement in Computers and Communications.
[4]. Han Lock Siew, Md Jan Nordin, “Regression Techniques for the Prediction of Stock Price Trend”.
[5]. K.R Sekar, “Gold price estimation using multivariable model”, International Conference on Networks and Advances in
Computational Technologies (NetACT) July 2017.

www.ijesi.org 32 | Page
 Stock Price Trend prediction using Multiple Linear Regression
[6]. Yahoo Finance Developer Network https://developer.yahoo.com/yql/?guccounter=1Apple Inc. A technology company
https://www.apple.com/Nasdaq Financial Company https://www.nasdaq.com/“Padas” Library documentation for the function
“fillna()” https://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.fillna.html
[7]. Lucas Nunno, “Stock Market Price Prediction Using Linear and Polynomial Regression Models”, 2014.
[8]. yale.edu, “Multiple Linear regression”, http://www.stat.yale.edu/Courses/1997-98/101/linmult.htm
[9]. Khaled Assaleh , Hazim El-Baz , Saeed Al-Salkhadi, “Predicting Stock Prices Using Polynomial Classifiers: The Case of Dubai
Financial Market ”, Journal of Intelligent Learning Systems and Applications, 2011, 3, 82-89
[10]. Phua, P.K.H. Ming, D., Lin, W. (2000), Neural Network With Genetic Algorithms For Stocks Prediction, Fifth Conference of the
Association of Asian-Pacific Operations Research Societies, 5th - 7th July, Singapore. J. Moody. Learning to trade via direct
reinforcement. IEEE Transactions on Neural Networks
[11]. A.S. Weigend and N.A. Gershenfeld, “Time Series Prediction: Forecasting the Future and Understanding the Past,” Addison-
Wesley, Reading, MA, 1993.
[12]. G. Grudnitzky and L. Osburn, “Forecasting S&P and Gold Futures Prices: An Application of Neural Networks,” Journal of Futures
Markets, vol. 13, No. 6, pp. 631-643, September 1993.
[13]. P.J. Werbos, “Generalization of back propagation with application to a recurrent gas market model,” Neural Networks, vol. 1, pp.
339-356, 1988.
[14]. Hiransha M , Gopalakrishnan E.A , Vijay Krishna Menon, Soman K.P, “NSE Stock Market Prediction Using Deep-Learning
Models”, International Conference on Computational Intelligence and Data Science (ICCIDS 2018)

Shruti Shakhla "Stock Price Trend Prediction Using Multiple Linear Regression "International
Journal of Engineering Science Invention (IJESI), vol. 07, no. 10, 2018, pp 29-33

www.ijesi.org 33 | Page

View publication stats