TABLE OF CONTENTS

CHAPTER ONE.. 1

INTRODUCTION.. 1

1.1      Background of the Study. 1

1.2      Statement of the Problem.. 2

1.3      Aims and Objectives. 3

1.4      Significance of the Study. 3

1.5      Scope of the Study. 4

1.6      Project Organization. 4

CHAPTER TWO.. 5

LITERTATURE REVIEW... 5

2.1      Background of the Study. 5

2.1.1       AI and Predictive Techniques. 6

2.2      Machine Learning. 8

2.2.1       Supervised Learning. 8

2.2.2       Unsupervised Learning. 9

2.2.4       Support Vector Machines. 10

2.3      Deep Learning and Artificial Neural Networks. 11

2.3.1       Recurrent Neural Network. 12

2.3.2       Long-short term memory. 13

2.3.3       Convolutional Neural Network (CNN) 14

2.4      Related Work. 15

CHAPTER THREE.. 17

METHODOLOGY.. 17

3.1      Introduction. 17

3.2      System Development Methodology. 18

3.2.1       Convolutional Neural Network (CNN) 18

3.2.2       LSTM (Long Short-Term Memory) 19

3.2.3       Bi directional LSTM... 20

3.2.4       Bidirectional LSTMs. 21

3.3      Dataset Description. 22

3.3.1       Proposed System Requirements. 22

3.3.2       Proposed System Design. 22

3.4      Software and Hardware Requirement 24

3.4.1       Hardware Requirement 24

3.4.2       Software Requirement 24

CHAPTER FOUR ………………………………………...……………………25

IMPLEMENTATION AND TESTING……………………...………..25

4.1     Introduction…………………………………………………………………………25

4.2     Choice of Programming Language……………………………………….25

4.3     Main Implementation……………………………………………………………….25

4.4     Simulation Results and Discussion………………………………………….27

4.4.1  Dataset Description………………………………………………...…27

4.4.2  Performance Metrics…………………………………………………………28

4.5     Performance Evaluation of Models based on CNN and BiLSTM…28

CHAPTER FIVE ……………………………………………………………….30

SUMMARY, CONCLUSION AND RECOMMENDATION………30

5.1     Summary…………………………………………………………………………....30

5.2     Conclusion………………………………………………………………………….30

5.3     Recommendation…………………………………………………………………....30

References. 32

APPENDIX (Source Code) ………………………………………………….36

CHAPTER ONE

INTRODUCTION

1.1   Background to the study

Stock is a financial product characterized by high risk, high return and flexible trading, which is favored by many investors. Investors can get abundant returns by accurately estimating stock price trends. However, the stock price is influenced by many factors such as macroeconomic situation, market condition, major social and economic events, investors’ preferences and companies’ managerial decisions. Therefore, prediction of the stock price has always been the focus and difficult research topic. Statistical and econometric models are generally used in traditional stock price prediction, but these methods cannot deal with the dynamic and complex environment of the stock market. Since 1970, with the rapid development of computer technology, researchers have begun using machine learning to predict stock prices and fluctuations, helping investors determine investment strategies to reduce risk and increase returns.

The stock market is a highly complex time series scenario and has typical dynamic characteristics. There will be a lot of stock dynamic trading after the opening of the market and stock price will change accordingly. Moreover, the stock price is affected by many unpredicted factors, which results in a typical nonstationary stock price time-series data. Therefore, stock price prediction is one of the most challenging problems in all kinds of prediction research. In the past decades, scholars have studied stock price prediction from many perspectives, where the improvement of prediction models and the selection of model features are the two most important directions among them. Most of the early studies used econometric models, such as autoregressive integrated moving average (ARIMA) and autoregressive conditional heteroskedastic autoregressive integrated moving average (ARCH-ARIMA) (Booth et al., 2001), to predict stock price. However, it is difficult for econometric models to consider the impact of other factors on stock price fluctuations and they have strong assumptions about the data, which are often difficult to meet (Le and Xie, 2018). Therefore, machine learning has been widely used in stock price prediction in recent years and many more suitable models for stock prediction have been proposed. Many studies have shown that deep learning has superior efficiency than other models (Marmer, 2008) and neural network models excel regression and discriminant models (Refenes et al., 1994). In terms of feature selection, some scholars explore the correlation between new features and stock price and some new features, including political factors, macroeconomic factors and investors’ sentiment, etc., have been incorporated into the prediction model (Cervello-Royo et al., 2015).

Predicting the accurate stock price has been the aim of investors ever since the beginning of the stock market. Millions of dollars’ worth of trading happens every single day, and every trader hopes to earn profit from his/her investments. Investors who can make right buy and sell decisions will end up in profits. To make right decisions, investors have to judge based on technical analysis, such as company’s charts, stock market indices and information from newspapers and microblogs. However, it is difficult for investors to analyze and forecast the market by churning all this information. Therefore, to predict the trends automatically, many Artificial Intelligence (AI) techniques have been investigated. (Vanaga and Sloka, 2020). Some of the first research in prediction of stock prices dates back to 1994, in which a comparative study (Sousa et al., 2019), with machine learning regression. models were performed. Since then, many researchers were investing resources to devise strategies for forecasting the price of the stock.

1.2         Statement of the Problem

Financial analysts investing in stock market usually are not aware of the stock market behavior. They are facing the problem of trading as they do not properly understand which stocks to buy or which stocks to sell in order to get more profits. In today’s world, all the information pertaining to stock market is available but analyzing all this information individually or manually is difficult due to the sporadic data.  Automation of the process is required to achieve the solution to the stock trading problem. Many technologies and techniques have been deployed with achievable results to help investors in analyzing stock exchange trading. To automate the understanding and analysis of such numerical time series, machine learning might provide a good result with intelligent decision. This will allow financial analysts to foresee the behavior of the stock that they are interested in and thus act accordingly.

1.3         Aim and Objectives

The aim of this project is to design and implement a deep learning model that can predict stock exchange

The objectives are; prediction models for stock markets, which is also the scope of this survey.

i.                    To make a literature search and identify algorithms to use for stock exchange prediction.

ii.                  To build a prediction model for stock markets.

iii.                To evaluate the performance of the built model

   1.4              Significance of the Study

Stock exchange prediction is very tedious for financial analyst due to huge number of information to understand and apply in order to gain profit in the investment. Despite that it consumes lots of time, dealing with volume of information always leads to inaccuracy.

This project work will add to the literature on the stock exchange prediction using Artificial Intelligence (AI) technique. Extracting and analysing huge number of information for the prediction by human is always prone to errors, whereas applying the AI technique appropriately not only produce accurate result, but also saves time and energy. Above all, more precision and timeliness are always achieved with the AI techniques applied.

    1.5         Scope of the Study

The scope of this research focused on financial analysis of stock exchange market to predict the price of stock exchange based on 10-year dataset obtained from TATA GLOBAL BEVERAGES LIMITED.

1.6         Project Organization

Chapter One introduces the study of stock market, machine learning in line with the project topic, including background of the study, statement of the problem, aims and objectives, significance of the study and project organization. Chapter Two deals with literature review of stock market, machine learning, support vector machine, deep learning, Convolutional Neural Network (CNN) and Long Short Time Memory (LSTM) and highlighted related works were discussed. Chapter Three presents the methodology adopted for ensuring and carrying out this project and the development of a model that predicts a dataset of stock market. Chapter Four presents the results and discussions of the models used in the prediction of stock market while Chapter Five provides the summary of the results, recommendations, and conclusion.