1

Artificial Intelligence-Based Inspection Of Stock

Market Data And Future Stock Prediction

A Capstone Project Report

Submitted in the partial fulfilment of the requirements for
the award of the degree of

Bachelor of Technology in
Department of Computer Science Engineering
by
2000039005 - Muhammad Azeez Ur Rahaman
2000039006 - Muhammad Safoora Begum
2000039014 – Jaddu Naga Sai Krishna
2000039007 - Arava Teja Mani Raju

under the supervision of

Dr. Ravi Kumar Tata

(Associate professor)

Department of Computer Science Engineering

K L E F, Green Fields,
Vaddeswaram- 522502, Guntur (Dist), Andhra Pradesh, India.
November,2023

2
 Declaration

The (Capstone Project) Report entitled “ Artificial Intelligence-Based Inspection Of

Stock Market Data And Future Stock Prediction “ is a record of bonafide work of Muhammad
Azeez Ur Rahaman, Muhammad Safoora Begum, Jaddu Naga Sai Krishna, Arava Teja Mani
Raju, submitted in partial fulfilment for the award of B.Tech in Computer Science Engineering
to the K L University. The results embodied in this report have not been copied from any other
departments/University/Institute.

Muhammad Azeez Ur Rahaman (2000039005)

Muhammad Safoora Begum (2000039006)
Jaddu Naga Sai Krishna (2000039014)
Arava Teja Mani Raju (2000039007)

3
 Certificate

This is to certify that the (Capstone Project) Report entitled “ Artificial Intelligence-
Based Inspection Of Stock Market Data And Future Stock Prediction” is being
submitted by Muhammad Azeez Ur Rahaman, Muhammad Safoora Begum, Jaddu
Naga Sai Krishna, Arava Teja Mani Raju submitted in partial fulfilment for the
award of B. Tech in Computer Science Engineering to the K L University is a record
of bonafide work carried out under our guidance and supervision.
The results embodied in this report have not been copied from any other
departments/ University/Institute.

Signature of the Supervisor

Signature of the HOD Signature of the External Examiner

4
 ACKNOWLEDGEMENT

We are greatly indebted to our K L Deemed to be a university that has provided a

healthy environment to drive me to achieve my ambitions and goals. I would like
to express my sincere thanks to my project in charge, for the guidance, support, and
assistance they have provided in completing this project. We express my sincere
gratitude and thanks to my project guide Dr. Ravi Kumar Tata and Dr. Swain for
his novel association of ideas, encouragement, appreciation, and intellectual zeal
which motivated us to venture into this project successfully. We express sincere
gratitude to our Coordinator Dr. Ravi Kumar Tata And Dr. Swain for his
leadership and constant motivation provided in the successful completion of our
academic semester.We record it as my privilege to deeply thank our pioneer Dr.
A . S e n t h i l HOD- CSE for providing us the efficient faculty and facilities to make
our ideas into reality.Finally, It is pleased to acknowledge the indebtedness
to all those who devoted themselves directly or indirectly to make this project
report success.

5
 ABSTRACT
This expansive exploration lodgings into the field of AI- driven stock request vaticination,
combining rulings from several methodical inquiries. The primary thing was to prize applicable
data from current exploration to support guide more prosperous investing styles. utilizing
prescribed keywords, a complete hunt across Scopus and trap of Science databases generated a
huge corpus of 69 titles, climaxing in a detailed valuation of 43 methodical reviews containing
over 379 papers. From this voluminous pool, ten reviews were precisely taken for addition in the
final dataset.Certain tendencies sat out among the exploration rulings. specially, brace vector
motors( SVM), long short- tenure mind( LSTM), and artificial neural networks( ANN) have
surfaced as the most dominant AI approaches in stock request vaticination sweats. likewise,
literal ending stock prices surfaced as the primary data source, featuring their overcritical
significance in modeling prophetic fabrics. likewise, the common emphasis on delicacy as the
ultimate interpretation criterion persisted throughout the inquiries, establishing its overcritical
part in arbitrating prophetic efficacity. Despite the uproariousness of perceptivity gained,
significant exploration hiatuses and implicit rows for inquiry surfaced. specially, the eventuality
for invention rests in the exploration of assorted data sources and their subtle amalgamations,
which pledge to ameliorate prophetic models with further inflexibility and adaptability. likewise,
the relative examination of miscellaneous AI methodologies and ways is an intriguing area ripe
for disquisition, as each methodology has distinct advantages and utility in nonidentical portions.
Eventually, the demand to ameliorate valuation procedures for vaticination pointers and norms
rings true, indicating the want for a further refined understanding of prophetic model efficacity
and real- world sequel. The stock request is a sprightly business where pieces of intimately listed
pots are bought and vended, with actors hoping to benefit from freight swings. Employing slice-
bite technology like artificial intelligence( AI) has enormous pledge for perfecting stock freight
cast delicacy. Our model aims to propose a new expressway to prognosticating stock motions
with high perfection. After precisely esteeming multitudinous ways and procurators, we
discovered that engine literacy( ML) algorithms similar as Random Forest, Long Short- Term
Mind( LSTM), Brace Vector motors( SVM), and Artificial Neural Networks( ANN) are
underutilized. We want to close this gap by employing these sophisticated algorithms for lesser
soothsaying interpretation. Our fashion is grounded on the addition of numerous datasets,
involving previous time's share request valuations, literal currency and goods request data, and
archival news captions. These datasets go through expansive medication to solace their
felicitousness for study. therefore, our path focuses a strong emphasis on the preprocessing step,
admitting its overcritical part in setting the root for operative vaticination. After preprocessing,
we apply significant AI algorithms to the culled datasets, with the thing of rooting applicable
perceptivity and driving productive effects. We want to discover untapped eventuality in stock
request vaticination by utilizing the capabilities of Random Forest, LSTM, SVM, and ANN,
accordingly adding the efficacity of soothsaying sweats. Likewise, our alluded path goes
beyond theoretical abstraction by laying its real- world operation and naming effects with
vaticination delicacy.

6
 CONTENTS
Page No
Abstract VII

CHAPTER-1
Introduction 8-10
CHAPTER-II
Literature Survey 11-14
CHAPTER-III
AI in stock market Introduction
3.1 what is AI in stock market? 15-17
3.2 Process involved in AI stock market? 18-22
3.3 Problem Description in AI stock market 23-24
3.4 Problem Methods in AI stock market 25-28
CHAPTER-IV
4.1 Experimental Investigations 29-30
4.2 Algorithm 30-36
4.3 Experimental Results 37-40
CHAPTER-V
Results & Analysis 41-42
Conclusion 42-43
References 43-48

7
 CHAPTER - I
INTRODUCTION

Vaticinating stock request trends is a daunting but necessary striving for investors, dealers,
and experimenters likewise. A mass of approaches, gauging fine, statistical, and Artificial
Intelligence( AI) fields, have risen to read stock valuations and outperform request prospects.
Engine literacy( ML) and Deep literacy( DL) ways in AI've entered adding concentration
for their implicit to revise stock request vaticination( Alshater etal., 2022; Chhajer etal., 2022;
Liang etal., 2020; Macchiarulo, 2018; Mokhtari etal., 2021; Samitas etal., 2020; Song & Jain,
2022; Sun etal., 2016). savants have delved numerous aspects of the integration of AI
technology with stock request soothsaying, feting the eventuality of cold-blooded ways to
ameliorate vaticination delicacy.In the interest of perfecting title, this work follows the
delineations given away by Jakhar and Kaur( 2020). Astronomically stated, AI includes
investing computers with mortal- suchlike intelligence, and any computer program that has
factors of mortal intelligence falls under this order. Engine literacy( ML), a subset of AI,
allows computers to get from data without unequivocal programming, with the thing of
furnishing motors with resolution- making chops grounded on reused data and information.
ML is basically a fashion for utilizing AI's capabilities in automated literacy and resolution-
making processes. Deep literacy( DL) develops as a subset of engine literacy( ML),
distinguished by the use of artificial neural networks( ANN) that mimic the complications of
natural neural networks discerned in the mortal brain. DL, analogous to the brain's cognitive
processes,Deep literacy( DL) develops as a subset of engine literacy( ML), distinguished by
the use of artificial neural networks( ANN) that mimic the complications of natural neural
networks discerned in the mortal brain. DL, like the brain's cognitive processes, organizes
and analyzes information by categorizing and labeling data points, performing in remarkable
delicacy, indeed with unshaped datasets. still, the success of DL is dependent on the vacuity
of voluminous volumes of data for training, pressing the resource- ferocious character of this
fashion. In summary, this deliberation emphasizes the revolutionary authority of AI, ML,
and DL in changing stock request cast styles. By exercising these ultramodern tools and
espousing cold-blooded ways, academics and interpreters hope to outclass old boundaries,
adding vaticination delicacy and unleashing new.The study composition" exercising Engine
mastering for Stock request soothsaying" produced by Raut Sushrut Deepak, Shinde Isha
Uday, andDr.D. Malathi emphasizes the overcritical significance of reaching high situations
of delicacy and perfection in stock freight vaticination sweats. Traditional approaches like
as time series dissection, abecedarian dissection, and specialized dissection are still
extensively exercised by investors and institutions, although their credibility is constantly
called into mistrustfulness. As a result, there's a satisfying want to probe other ways to
ameliorate stock request cast efficacity. In this exploration, the seat moves to exploiting
Engine literacy( ML) approaches, wherein models are trained on literal stock data to get
perceptivity and make,Jan Ivar Larsen scratched the composition" Stock Price Prediction

8
utilizing Technology Analysis and Mechanical Learning." tallying to the report, once stock
prices can expect unborn freight changes. The stock freight model employs a two- league
discussion fashion. The first subcaste uses specialized dissection to guide the alternate
subcaste, which is grounded on engine literacy. The model is supported by a fiscal operation
fashion that invests in unborn protrusions grounded on once prognostications' acquirement.
The cast model outperforms the Oslo Benchmark Index( OSEBX) grounded on several
portfolio simulations and trading styles.OmarS. Soliman, Osman Hegazy, and Mustafa Abdul
Salam's paper" An AI- Grounded Approach to Predicting Stock Market Trends" delves into
the demesne of share request freight vaticination, with the thing of soothsaying future
valuations of stock prices and other fiscal instruments traded across colorful exchanges. The
capability to duly prognosticate share prices has enormous implicit for adding individual
investor profitability. To break this conclusion, the exploration introduces a new engine
literacy model aimed simply for assaying share request freight motions. The alluded fashion
combines flyspeck mass optimization with arbitrary timber algorithms. specially, the
flyspeck mass optimization path is exercised to ameliorate the arbitrary timber model's
interpretation by chancing an ideal set of parameters.The algorithm is trained to directly read
diurnal prices by utilizing perceptivity from former share request statistics and specialized
pointers. This new fashion not only mitigates enterprises like traffic and original minima,
but it also greatly improves cast delicacy, making it an seductive path for investors appearing
to capitalise on assiduity trends."Forecasting Models for the Indian Stock Market" examines
the difficulty of projecting share market price patterns in an environment marked by constant
and massive data variations. Recognizing the stock market's intricacy and volatility, investors
negotiate a dynamic environment filled with financial rewards and losses. In answer to this
problem, the study seeks to construct predictive models for both short-term and long-term
price estimates.1.1.1. Lack of Clear Understanding There's a lack of clarity regarding the
extent to which CSR enterprise appreciatively or negatively affect organizational
performance within Ethio Telecom Jimma branches. This nebulosity may stem from varying
interpretations of CSR's part and its perceived impact on fiscal criteria , hand satisfaction,
client retention, and overall commercial image. . Resource Allocation Challenges Ethio
Telecom Jimma branches may face difficulties in allocating coffers effectively to CSR
conditioning while balancing the need for fiscal profitability. Limited fiscal coffers and
contending precedences may hamper the perpetration of comprehensive CSR strategies,
potentially affecting the company's capability to achieve its organizational objects. .
Stakeholder prospects Managing the different prospects of stakeholders, including guests,
workers, government agencies, and original communities, presents a significant challenge for
Ethio Telecom Jimma branches. Meeting the demands of these stakeholders while
contemporaneously addressing CSR pretensions and organizational performance targets
requires careful navigation and strategic alignment. . dimension and Evaluation The absence
of standardized criteria or evaluation fabrics to assess the impact of CSR enterprise on
organizational performance poses a challenge. Without clear marks or pointers, it becomes
grueling for Ethio Telecom Jimma branches to gauge the effectiveness of their CSR sweats
and make informed opinions about resource allocation and strategy refinement. . Regulatory
Compliance Ethio Telecom Jimma branches operate within a nonsupervisory frame that may
put fresh conditions or constraints related to CSR practices. icing compliance with these

9
regulations while contemporaneously pursuing organizational performance pretensions can
produce pressure and complexity for the company. Cultural and Societal Context The artistic
and societal environment in which Ethio Telecom Jimma branches operate may impact the
perceived significance and effectiveness of CSR enterprise. Understanding and navigating
these contextual factors is essential for aligning CSR sweats with organizational performance
objects and achieving meaningful impact.Guangyu Ding and Liangxi Qin's exploration
composition" Exploring Stock Price Prediction utilizing LSTM- Grounded Associated
Network Model" investigates the changing terrain of the stock request, which continues to
be a seat point for both individual investors and institutional pots. In this ever- changing
terrain, multitudinous procedures are exercised to expect unborn share valuations, with
separate orders for fine styles and artificial intelligence( AI) ways. Mathematical styles
carry methodical and bow models, whereas AI ways carry Multilayer Perceptron( MLP),
Convolutional Neural Network( CNN), Bayes Network, Backward Propagation Network,
Support Vector Machine( SVM), and Single- Subcaste Long Short- tenure Mind( LSTM).
Understanding and navigating these contextual factors is essential for aligning CSR sweats
with organizational performance objects and achieving meaningful impact.Guangyu Ding and
Liangxi Qin's exploration composition" Exploring Stock Price Prediction utilizing LSTM-
Grounded Associated Network Model" investigates the changing terrain of the stock request,
which continues to be a seat point for both individual investors and institutional pots.
Specially, . dimension and Evaluation The absence of standardized criteria or evaluation
fabrics to assess the impact of CSR enterprise on organizational performance poses a
challenge. Without clear marks or pointers, it becomes grueling for Ethio Telecom Jimma
branches to gauge the effectiveness of their CSR sweats and make informed opinions about
resource allocation and strategy refinement. . Regulatory Compliance Ethio Telecom Jimma
branches operate within a nonsupervisory frame that may put fresh conditions or constraints
related to CSR practices. icing compliance with these regulations while contemporaneously
pursuing organizational performance pretensions can produce pressure and complexity for
the company. the seat of this study is on the evolution of LSTM- grounded associated
network models, which enable the appraisal of colorful variable,It's needed to develop a
model that can vary A deep neural network model is developed to affair valuations in parallel
while taking multitudinous inputs. This short- tenure mind network has several inputs and
labors. This network can estimate share prices at numerous situations, involving open, high,
and low. Accessible network model assimilated to Long Short- tenure Mind and Deep Neural
Network models. The linked model outperforms other models in prognosticating several
variables in coincidently, with an estimated delicacy of over 95,(12).

10
 CHAPTER-II
LITERATURE SURVEY

The study" Machine Learning Approach in Stock Market Prediction" emphasizes the
significance of delicacy and perfection in prognosticating share prices, sweating the use
of time series dissection, abecedarian dissection, and specialized dissection by investors
and institutions. Feting the limitations of traditional approaches, the study calls for a
probative program grounded on Engine literacy( ML). Following significant algorithmic
study, the report indicates that Artificial Neural Network( ANN) is the stylish option, owing
to its adjustable nature with a wide range of variables and parameters. The efficacity of the
engine mastering ideas is estimated utilizing data from the Bombay Stock Exchange
indicator dataset.In discrepancy, Jan Ivar Larsen's work," Stock Price Prediction utilizing
Technology Analysis and Mechanical Learning," focuses on the use of former stock prices
to expect unborn motions. The model uses a two- league discussion system, incorporating
specialized dissection rulings into the first subcaste and guiding the alternate subcaste
utilizing engine literacy. likewise, the model uses a fiscal operation fashion grounded on
once cast interpretation to decide unborn investment appropriations. The exploration uses
portfolio simulations and trading models to illustrate the model's capacity to beat the Oslo
Benchmark Index( OSEBX).The paper delves into the demesne of stock request
vaticination, featuring the bid to estimate unborn valuations of stock prices and other traded
fiscal instruments across colorful exchanges. Feting the eventuality for swelled gains for
individual investors through prosperous soothsaying, the paper proposes a engine literacy
model acclimatized for this purpose. This proffered model amalgamates flyspeck mass
optimization and arbitrary timber algorithms to enhance vaticination delicacy. specially,
the flyspeck mass optimization algorithm is assumed to compound the arbitrary timber
frame, using once request datasets and specialized pointers. Through scrupulous parameter
election eased by flyspeck mass optimization, the model aims to alleviate expostulations
similar as original minima, thereby bolstering its prophetic capabilities.The handed model
is also strictly tried across several fiscal databases and assimilated to the Leuralberg-

11
Marquardt neural network ways. The effects of this comparison dissection punctuate the
alluded model's better delicacy and adaptability, which may be attributed to its operative
use of flyspeck mass optimization. The first number emphasizes the difficulty of bunching
stock valuations in a dynamic and daedal request terrain. It distinguishes two vaticination
midairs short- tenure and long- tenure vaticinations, and uses mastering algorithms to
make prophetic models. The composition emphasizes the integration of literal data and
feelings, as well as the great delicacy attained utilizing standardized engine mastering
approaches, specially in diurnal cast models. likewise, it investigates the commerce of
yearly patterns, indicating connections between posteriormonths.The alternate study
lodgings into the multidimensional character of stock request dynamics, which attracts a
wide range of parties, involving people and investment associations. It clarifies the
perfection between fine approaches and AI ways for soothsaying future share valuations,
arguing in indulgence of a deep neural network model able of recycling colorful inputs and
labors at formerly. flaunting great delicacy surpassingThis model, which has a 95 delicacy,
uses short- tenure mind networks to estimate several characteristics of share prices
contemporaneously.The third study focuses on the evolution of vaticination models that
exercise several procurators similar as oil painting prices, foreign trade classes, and news
feeds to expect request spirit as positive or inhospitable. utilizing a variety of engine
literacy algorithms, involving Brace Vector motors and intermittent Neural Networks, the
exploration emphasizes the significance of pattern matching in perfecting vaticination
delicacy. Eventually, the fourth study discusses the changing terrain of stock freight
vaticination, featuring the significance of news events in arbitrating request motions. It
emphasizes the measure from traditional logical approaches to automated models powered
by engine literacy algorithms. To more directly expect request motions, these models
exercise event- related news as well as share prices and indicators.The alluded path focuses
on utilizing several engine mastering algorithms to expect unborn stock valuations. The
system efficiently trains and evaluates several ML algorithms to read unborn share
valuations grounded on literal data. It specially uses End of Day( EOD) data from once
times to train and estimate engine literacy models. Data processing, visualization, and
model construction are carried out utilizing a variety of engine literacy libraries and fabrics,
similar as Numpy, Pandas, Sklearn, Keras, and TensorFlow. likewise, these libraries make

12
fine missions on data moreefficient.Furthermore, the system uses past news captions for
passion dissection, which improves the models' soothsaying dominions. The major archive
exercised is NumPy, which is well- known for its capability to manipulate and clean datasets,
allowing it to be exercised directly for ML models. SklearnThe archive is exercised for
factual computation, estimation, and vaticination missions. The dataset contains literal
share request data gathered from multitudinous public and open internet depositories, with
the brunt( 85 or further) exercised to train ML models and the remaining for confirmation
and testing.This dissection identifies major areas that have gotten inadequate erudite
disquisition and demand concentration. To begin, experimenters constantly disregard or
deficiently manage the election and treatment of input data, both of which play overcritical
places in AI models. The quality and content of input data have a direct jolt on time series
estimation and model training; nonetheless, this element is constantly overlooked.
Alternate, the delicacy of prognostications is explosively dependent on the liberty of
applicable parameters for erecting AI models. Authors constantly pay nonidentical
amalgamations of input variables and model parcels, which give good effects. still,
perfecting model project remains delicate, with numerous depending on trial- and-
inaccuracy approaches that lack putative argument. Likewise, judging prognosticating
delicacy simply on dissonances from factual valuations may be inadequate for fiscal
request interpreters seeking return maximization. Indeed modest vaticination crimes can
affect in respectable disadvantages in stock trading, sweating the want of good directional
soothsaying. Third, the value of datapre-processing in experimental ways is constantly
overlooked in current literature. operativepre-processing, similar as data scaling and
perceptivity dissection for index election, is overcritical for perfecting model vaticination
delicacy. common, prostrating these effects might vastly ameliorate the efficacity and
responsibility of AI- grounded stock request vaticination models.Stock request
vaticinations have colluded people for times, not simply out of curiosity, but also because
request changes have far- reaching worldwide profitable consequences. Over time, artificial
intelligence( AI), especially deep literacy models, has surfaced as a feasible path for
perfecting the delicacy of stock future indicator vaticinations. This special expressway to
soothsaying stock request futures utilizing artificial intelligence has the implicit to
significantly support stock dealers and investors. Despite the essential difficulty of studying

13
fiscal requests, advances in artificial intelligence have produced encouraging issues. AI
styles, involving Artificial Neural Network( ANN) and Swarm Intelligence,Algorithms like
Artificial Neural Network( ANN) and Swarm Intelligence may make incredibly accurate
prognostications. Recent trials on major stock exchanges in India, similar as the National
Stock Exchange( NSE) and Bombay Stock Exchange( BSE), have exercised emulsion
vaticination models grounded on AI ways similar as ANN, Random Forest( RF), and Brace
Vector Regression( SVR) to read unborn stock prices over a variety of time ages. Over
time, artificial intelligence( AI), especially deep literacy models, has surfaced as a feasible
path for perfecting the delicacy of stock future indicator vaticinations. This special
expressway to soothsaying stock request futures utilizing artificial intelligence has the
implicit to significantly support stock dealers and investors. Despite the essential difficulty
of studying fiscal requests, advances in artificial intelligence have produced encouraging
issues. AI styles, involving Artificial Neural Network( ANN) and Swarm
Intelligence,Algorithms like Artificial Neural Network( ANN) and Swarm Intelligence may
make incredibly accurate prognostications. The essential cases of stock request
vaticination, similar as high volatility, nonlinearity, and chaos, have driven the
objectification of abecedarian statistical ways, pointers, and improved AI algorithms.
likewise, the relationship between transnational and original requests has been delved in
order to exercise engine literacy to ameliorate stock request prognostications, with the
thing of outperforming request indicators. Experimenters remain to probe innovational
ways, similar asThe argentine vaticination system and mongrel models are exercised to
boost stock cast delicacy, especially in bullish or bearish request portions. These
advancements punctuate the significance of algorithms and engine literacy in enhancing
stock request responses.

14
 CHAPTER -III

3.1 What is AI IN STOCK MARKET?

Artificial intelligence(AI) has the implicit to revise several spots,
involving the stock request. At its heart, artificial intelligence(AI) is the coinage of
computer systems able of serving conditioning that would typically need mortal
intelligence. In the stock request, AI systems dissect voluminous volumes of data, discover
trends, and induce vaticinations to support investors make opinions. Prophetic analytics
is one of the most common or garden operations of artificial intelligence in the stock request.
To estimate unborn stock prices or request trends, AI models can exercise engine literacy
algorithms to charge literal request data, commercial financials, profitable pointers, and
indeed news passion. These prophetic algorithms enable investors to expect request
oscillations and make better educated trading opinions. Another important use of AIIn the
stock request, AI is exercised for algorithmic trading, frequently known as high- frequence
trading( HFT).

AI- powered trading algorithms make deals grounded on destined parameters, similar as
freight changes, measure variations, or specialized suggestions. These algorithms can
estimate massive volumes of data in real time and apply deals at classes far above mortal
capabilities, allowing dealers to pinch deciduous request openings. Passion dissection is
another area where AI has a huge jolt in the stock request. Artificial intelligence systems

15
may charge request passion and investor sentiments by assessing news stories, gregarious
media dispatches, and other textual data. This information assists dealers in gathering
request passion patterns, relating practicable request carriers, and awaiting oscillations in
investor spirit that may prompt stock prices. threat operation is another important point
of AI in the stock request. AI- powered threat operation systems look at portfolio
composition, request volatility, and other threat pointers to identify practicable pitfalls
and reduce inhospitable threat. These systems use daedal statistical models and engine
literacy algorithms to estimate threat and enhance portfolio interpretation while guarding
capital. common, the objectification of AI in the stock request has converted the
expressway investorsApproaching trading and portfolio operation. By using the authority
of AI- driven analytics, dealers may gain deeper perceptivity into request dynamics, induce
more accurate vaticinations, automate trading procedures, and take pitfalls more efficiently,
performing in better investment issues.Artificial intelligence( AI) is one of the most
disruptive and instigative motifs in ultramodern technology.

AI is unnaturally concerned with the coinage of computer systems able of executing

conditioning that bear mortal intellect, similar as literacy, logic, case working, perception,
and language understanding. While the notion of AI has senior wells, organized exploration
and evolution as a scientific area began in themid-twentieth century. Computer scientist
John McCarthy began the expression" artificial intelligence" in 1956, during the Dartmouth
Conference, when he alluded the coinage of robots able of bluffing mortal intellect.
McCarthy, along with other settlers like Allen Newell, HerbertA. Simon, and Marvin
Minsky, established the foundation for AI exploration and development. AI creations have
redounded in a wide range of operations, revamping diligence and changing the

16
expressway we reside and work. Some of the primary advantages of AI carry robotization
Artificial intelligence enables the robotization of repetitious and monotonous missions,
discharging up mortal coffers for further innovational and strategic enterprise. AI- driven
robotization improves effectiveness and productivity in a variety of operations, involving
robotic process robotization( RPA) in commercial missions and independent buses in
transportation. resolution Making AI systems can estimate voluminous volumes of data,
descry trends, and make data- driven judgments snappily and directly. AI- powered
resolution brace systems help professionals in diligence similar as finance, healthcare, and
fabricating in making informed opinions and optimizing effects. Personalization AI
systems may charge stoner preferences, geste, and literal data to give
acclimatizedexperiences.and suggestions. AI improves consumer happiness and
engagement in a variety of settings, involving substantiated product suggestions one-
commerce platforms and acclimatized content on streaming services. Prophetic Analytics
AI allows for prophetic modeling and soothsaying by studying literal data and detecting
trends and patterns. Prophetic analytics operations carry bunching client development,
deals trends, stock prices, and rainfall patterns. Advanced effectiveness and Cost Savings
By automating procedures, perfecting resource allocation, and simplifying missions, AI
assists enterprises in adding effectiveness and cost savings. AI- driven results may exclude
miscalculations, waste, and enhance resource use, performing in respectable cost savings
over time. common, AI has enormous pledge for working convoluted cases, prodding
invention, and arbitrating the future of technology. still, it also raises ethical, artistic, and
nonsupervisory enterprises that bear

17
3.2 Process involved in AI stock market?

• Data Collection
• Data Preprocessing
• Feature Selection
• Data Cleaning
• Feature Selection
• Model Selection
• Training the Model
• Cross-Validation

3.2.1 Data Collection

Gather historical stock market data from various sources such as financial databases, APIs, or
online repositories. This data typically includes information like stock prices, trading volumes,
market indices, company financials, news articles, and social media sentiment.Data collecting is a
overcritical element of AI- powered stock request dissection, allowing computers to make
accurate vaticinations and judgments. In the environment of AI in the stock request, data
collecting entails acquiring, recycling, and assaying massive volumes of data from numerous
sources in order to discover patterns, trends, and signals that may be exercised to guide trading
strategies. Allow's go into the medications of data collecting in AI stock request operations.
Market Data This contains literal pricing data, trade volumes, shot- interrogate spreads, and other
request parameters for colorful fiscal instruments similar as justice, bonds, goods, and
derivations. request data serves as the foundation for assaying literal request exertion and
discovering trends that support prognosticate unborn freight motions. Abecedarian data refers
to information on the fiscal health and interpretation of special companies.

18
3.2.2 Data Preprocessing

Clean the collected data by handling missing values, removing outliers, and standardizing the
format. This step also involves feature engineering, where you create new features from the
existing data to improve the predictive power of the model.Select the most relevant features that
are likely to influence stock prices. This can be done using techniques like correlation analysis,
statistical tests, or domain knowledge.Data preprocessing plays a crucial role in AI-driven stock
market analysis, as it involves cleaning, transforming, and preparing raw data to make it suitable
for use in machine learning algorithms. Here's an overview of the data preprocessing steps.

3.2.4 Data Cleaning

The first step in data preprocessing is to clean the raw data by removing any errors, inconsistencies,
or missing values. In stock market data, missing values may occur due to gaps in trading days or
incomplete data feeds. Techniques such as interpolation, imputation, or simply removing
incomplete records can be used to handle missing data.Normalization and Scaling: Stock market
data often consists of features with different scales and ranges. Normalization and scaling
techniques are used to bring all features to a similar scale, preventing certain features from
dominating others during the model training process. Common techniques include Min-Max
scaling, Z-score normalization, and robust scaling.Feature Engineering: Feature engineering
involves creating new features or transforming existing ones to enhance the predictive power of
the data. In the context of stock market analysis, this may involve calculating technical indicators
such as moving averages, exponential moving averages, relative strength index (RSI), and MACD
(Moving Average Convergence Divergence). These technical indicators can provide valuable
insights into market trends and momentum.Handling Categorical Data: In stock market data,
categorical variables such as sector, industry, or stock exchange may need to be encoded
numerically before they can be used in machine learning algorithms. Techniques such as one-hot
encoding or label encoding can be used to convert categorical variables into numerical
representations.

19
3.2.5 Feature Selection

Feature selection involves identifying the most relevant features that contribute to the predictive
power of the model while discarding irrelevant or redundant features. This helps reduce the
dimensionality of the data and improves the efficiency and interpretability of the model.
Techniques such as correlation analysis, feature importance ranking, and model-based selection
can be used for feature selection.Time Series Handling: Stock market data is often in the form of
time series, with observations recorded at regular intervals (e.g., daily, hourly). Time series-
specific preprocessing techniques may include resampling to a uniform time interval, handling
seasonality and trend components, and splitting the data into training and validation sets while
preserving temporal dependencies. By performing these data preprocessing steps, analysts and
data scientists can prepare high-quality datasets that are suitable for training AI models to make
accurate predictions and informed decisions in the stock market. Effective data preprocessing is
essential for ensuring the reliability and robustness of AI-driven trading strategies.

20
3.2.6 Model Selection

Choose the appropriate AI model for stock market prediction. Commonly used models include
Regression models Linear Regression, Ridge Regression, Lasso Regression Tree-based models
Decision Trees, Random Forest, Gradient Boosting Machines Neural network models Artificial
Neural Networks (ANN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM)
networks Ensemble methods Voting, Bagging, Boosting.

3.2.7 Training the Model

Split the preprocessed data into training and testing sets. Train the selected AI model on the
training data using supervised learning techniques. The model learns the underlying patterns and
relationships between the input features and the target variable (e.g., stock prices) Model
Evaluation Evaluate the trained model's performance on the testing data using appropriate
evaluation metrics such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), or
Mean Absolute Percentage Error (MAPE). This step helps assess the model's accuracy and
generalization ability Hyperparameter Tuning Fine-tune the model hyperparameters to optimize
its performance. Hyperparameters are configuration settings that control the learning process of
the model, such as learning rate, number of layers, and number of neurons in a neural network.

21
3.2.8 Cross-Validation

Perform cross-validation to validate the model's performance across multiple subsets of the data.
This helps ensure that the model's performance is consistent and reliable across different data
samples.Prediction and Deployment: Once the model is trained and validated, use it to make
predictions on new, unseen data. Deploy the trained model into production or integrate it into
trading platforms for real-time predictions.Monitoring and Updating: Continuously monitor the
model's performance in the live environment and update it periodically to adapt to changing market
conditions. This may involve retraining the model with new data or adjusting its parameters based
on feedback.By following these steps, you can develop and deploy AI-powered models for stock
market prediction that help investors make informed decisions and maximize returns.

22
3.3 Problem Description in AI stock market
Applying artificial intelligence(AI) to the stock request faces a physical
challenge known as request inefficiency. fiscal requests are largely daedal and dynamic
ecosystems impacted by a variety of variables, involving profitable data, geopolitical
events, and investor spirit. Despite the emergence of improved AI algorithms for awaiting
request motions, getting harmonious and predictable responses remains delicate owing to
the essential inefficiencies and unpredictability of fiscal requests. The difficulty of
effectively assaying and vaticinating mortal is a crucial element of the request
inefficiency dilemma. While AI algorithms are excellent at recycling voluminous volumes
of data and feting patterns, they constantly struggle to capture the nuances of mortal
resolution- timber. feelings, prejudices, and immoderation are all aspects that impact mortal
request geste. Likewise, fiscal data contains bruit and unpredictability, which complicates
the job of AI systems. Stock prices change due to a variety of variables, involving academic
trading, arb request bruit,and exogenous impacts. Distinguishing factual request signals
from arbitrary bruit, is a crucial conclusion for AI systems, which must stir through
massive volumes of data to induce accurate vaticinations. likewise, the case of overfitting
poses a significant challenge to the coinage of believable AI models for stock request
vaticination. Overfitting happens when a model learns to fit the bruit.in the training data
preferably than the underpinning patterns, which results in penurious conception to new
data. Overfitting is a ubiquitous peril that must be ruled in the convoluted and
unpredictable fiscalmarkets.attentively to solace the robustness and responsibility of AI-
driven trading ways. Likewise, the volume and quality of data give physical obstacles in
constructing effective AI models for stock request vaticination. fiscal data is constantly
disintegrated, inaccurate, and apt to miscalculations, making it delicate to induce secure
datasets for training AI systems. literal data may fail to represent the complications and
nuances of real- world request dynamics, limiting the effectiveness of AI models in
soothsaying future trends. Another major difficulty is the necessity for ongoing adaption
and literacy in reaction to changing request conditions. fiscal requests are represented by
changing trends, shifting investor preferences, and unanticipated circumstances, which
bear AI algorithms to acclimatize and evolve over time. Failure to reply to changing request
conditions can affect in obsolete models and penurious trading tactics, emphasising the

23
necessity ofcontinuing research and development in artificial intelligence for stock market
prediction. Eventually, the conclusion of translucency and interpretability is a physical
chain in espousing AI- powered trading systems. numerous AI systems, especially deep
literacy models, are generally regarded as black boxes, making it delicate to explain their
conclusions and comprehend the logic behind their prognostications. A lack of
translucency can corrode trust and confidence in AI- powered trading systems, generating
worries among investors, controllers, and other stakeholders. To epitomize, diving the
dilemma of request inefficiency in stock request AI entails prostrating technological,
ethical, legit. and ultrapractical walls. Developing strong, reliable, and transparent AI
algorithms able of conning the complications and misgivings of fiscal requests is an
instigative field of exploration and evolution in the fiscal region. Despite the obstacles, the
implicit vantages of AI- driven trading systems.A crucial aspect of the request inefficiency
case lies in the difficulty of directly modeling and prognosticating mortal geste . While AI
algorithms exceed at assaying voluminous datasets and relating patterns, they frequently
struggle to capture the nuances of mortal resolution- timber, which can be told by feelings,
impulses, and immoderation. As a result, AI models may fail to expect unforeseen shifts in
investor passion or request passion, leading to sour trading opinions. Another significant
challenge is the presence of bruit and randomness in fiscal data. Stock prices are subject to
oscillations caused by a variety of procurators, involving arbitrary request, academic
trading, and foreign impacts. Distinguishing between genuine request signals and arbitrary
bruit poses a redoubtable challenge for AI algorithms, which must filter out inapplicable
information to make accurat.prognostications.

24
3.4 Problem Methods in AI stock market
To get a thorough understanding of the methodology and reporting norms exercised in
methodical literature reviews in the field of stock request vaticination, we strictly examined
current inquiries accessible in the public sphere. This attempt rigorously followed the
regulations established by the PRISMA 2020 Statement involving both its roster and inflow
illustration to solace methodological robustness. likewise, we exercised the perceptivity
offered by the PRISMA Elaboration and Explanation to ameliorate clarity and appreciation of
the PRISMA norms.Our evaluation comprehended a thorough examination of published
methodical inquiries on folks and the perpetration of AI technologies.

Kaplan and Haenlein's( 2019) description of AI punctuated the system's capacity to read
foreign inputs, get constantly, and acclimatize flexibly to fulfill preset pretensions. We
primarily named inquiries that appeared at justice via the lens of stock request dynamics or
stock freight vaticinations. Utilizing the PRISMA 2020 guidelines( runner, McKenzie, etal.,
2021) as a companion, we strictly searched two major databases, Scopus and trap of Science,
for applicable methodical reviews on stock request vaticination utilizing AI technologies
similar as Engine literacy( ML) and Deep literacy( DL) styles. strict addition and rejection
criteria were exercised to stir through the recaptured commentaries, performing in the
election of 10 publications for full examination. The named estimations were also uprooted
and enciphered, allowing for in- depth examination across important confines similar as
vaticination methodology, data sources, interpretation criteria , limitations, and unborn
suggestions.In extension to following a painstakingly aimed exploration methodology, the
election of primary papers for this meta- review needed adherence to certain criteria. Drawing
on Kitchenham and Charters'( 2007) paradigm, our system comported of three separate
phases meticulous planning, rigid perpetration, and detailed dissection of effects. These
stages were overcritical to guaranteeing the validity and trustability of our rulings in the field
of methodical literature reviews for stock request vaticination.

As a result, the original step of this study comprises replying multitudinous essential
questions when the methodical review is completed, as well as defining the criteria for
addition, rejection, and quality evaluation. Table 1 outlines the exploration effects that will
serve as the seat of this methodical review. tallying to Jakhar and Kaur( 2020), the addition(
IC) and rejection( EC) criteria are determined by parameters similar as training data quantum
and needed tackle and software coffers.Despite the multitudinous methodical reviews
synthesizing statistical rulings on the crossroad of artificial intelligence( AI) and stock freight
vaticination, there's still a clear lack of agreement on the most operative and applicable AI
ways for this sphere, as well as their relative efficacity against indispensable methodologies.
This conclusion, reflected by Makridakis, Hyndman, and Petropoulos( 2020), emphasizes the
significance of a refined knowledge attained from earlier methodical exploration. specially,
Pinto, Figueiredo, and Garcia( 2021) set up significant exploration hiatuses, sweating the
significance of reconsidering our path to AI and stock request dynamics utilizing

25
assignments from former methodical reviews. New information has a daedal jolt on stock
pricing dynamics, a miracle that usual stock request breakdowns don't regard for, as
explained by two seminal fiscal propositions the Random Walk Model( Fama, 1995; Fama,
Fisher, Jensen, & Roll, 1969) and the Effective request thesis( Fama, 1965). Despite the
abecedarian premise of these propositions, a number of experimenters have challenged their
hypotheticals, establishing the potentiality of partial request pungency grassed by
socioeconomic proposition and perceptivity from behavioral economics and finance( Chong,
Han, & Park, 2017; Oliveira, Cortez, & Areal, 2017; Patel, Shah, Thakkar, & Kotecha, 2015;
Weng, Ahmed, & Megahed, 2017).

Recent inquiries show off a significant discrepancy in the methodology exercised for stock
request dissection, which are separated into fine and artificial intelligence( AI) strategies.
Mathematical ways generally exercise statistical tools, whereas AI styles largely exercise
engine literacy( ML) algorithms, as explained. specially, a major quantum of current
exploration uses engine literacy algorithms to estimate stock request vaticination capability,
establishing their ubiquity in the area. likewise, as a subset of AI, ML includes deep literacy(
DL) ways, as stated by Jakhar and Kaur( 2020). As a result, the geography of stock freight
vaticination approaches may be previously codified into archetypal ML algorithms, DL
styles, and neural networks( NN). still, the variety of archetypal ML, DL, and NN
approaches.In reaction to the expanding discussion around AI and stock request vaticination,
this study proposes a special path a meta- review that totally checks methodical reviews to
discover recent advances in the area. This meta- review, which covers several firmaments
similar as economics, statistics, finance, and computer wisdom, uses specialized methodology
to fully grasp the assorted range of approaches to AI and stock requestresearch.This design
intends to help the project, program, and prosecution of unborn exploration in the motive
by strictly reviewing, likening, and recapitulating data from former methodical reviews.
Eventually, the abecedarian thing of this meta- review is to give useful information for
developing unborn stock request investment styles.

This will be fulfilled by relating true AI algorithms for stock request vaticination, outlining
applicable data sources from being exploration, and assaying extensively habituated success
pointers. Likewise, we want to identify hiatuses and limitations within being methodical
reviews and propose results to ameliorate the quality and rigor of unborn estimations.
Through this work, we want to identify areas of consensus and dispute among being
methodical reviews, in order to understand presumptive accounts for any antithetical effects
and conclusions. likewise, the purpose of this dissection is to exfoliate light on developing
trends and expostulations in the field of AI and stock request vaticination, as well as to
propose prospective lines and prospects for unborn exploration.Our meta- review includes a
variety of reviews that concentrate on stock request vaticination, especially in tours of stock
freight or return. These inquiries exercise a wide range of AI approaches, similar as engine
literacy( ML), deep literacy( DL), neural networks( NN), brace vector motors( SVMs), and

26
passion dissection. still, we specially count duties that concentrate on other aspects of the
stock request, similar as volatility, threat, portfolio optimization, and trading tactics. also,
reviews that exercisenon-AI methodology, similar as fine or statistical approaches, for stock
request vaticination are neglected from our exploration.

Algorithms similar as Artificial Neural Networks( ANN), Neural Networks( NN), and Brace
Vector Regression( SVR) have been extensively exercised to estimate stock request share
prices. ANN and NN are the most frequently exercised strategies, counting for a sizable
proportion of application in 30 exploration( Kumar, Sarangi, & Verma, 2022b). Traditional
engine literacy( ML) styles have been shown off to produce worse effects when assimilated
to deep literacy( DL) approaches; nevertheless, DL still confronts effects in examining
streaming data, allotted computing, and computer scalability. likewise, the use of convoluted
ML approaches, involving ensemble models and DL, has swelled in fashionability, with
ensemble models strutting significant prophetic authority overrunning other methodologies
similar as SVM and ANN.In the current frame, a different batch of algorithms are exercised
for stock request soothsaying, categorised into ranges similar as Autoregressive( AR),
Moving moderate( Mama ), Autoregressive Integrated Moving moderate( ARIMA), and
Autoregressive Moving moderate( ARMA), as well as inharmonious models like
Autoregressive tentative Heteroskedasticity( bow), Generalized Autoregressive tentative
Heteroskedasticity( GARCH), and Neural Network. also, AI fabrics similar as Naive Bayes,
k- Nearest Neighbors( k- NN), Brace Vector Machine( SVM), Linear Retrogression, Artificial
Neural Network( ANN), and Random Forest are exercised to ameliorate vaticination models.
These models are calibrated utilizing literal stock data.

Still, present-day models substantially calculate on a single algorithm for vaticination, failing
to combine the rulings of nonidentical algorithms or take into account volition algorithmic
ways for more accurate prognostications. former exploration indicates that standard-issue
divisions are inadequate for stock freight computation. likewise, the being system's
interpretation degrades as the operating terrain changes since it doesn't regard for foreign
events similar as news events or variables impacting pricing in other requests similar as the
Forex and reality requests. The use of a single data source brings bias into the system.
likewise, the current system lacks integration across input sources, challenging a restatement
step. likewise, the system generally depends on either literal data or media dissection alone,
forgetting practicable community between the two sets.The system is structured into distinct
modules, each serving a specific function:

• Data Collection: Information is gathered from diverse sources like Yahoo Finance and
Google Finance, primarily in .csv format. Additionally, the system employs the Google News
API for acquiring news data.

• Analysis Manipulation and Visualization of Data: Collected data undergoes cleaning and
preprocessing to ensure suitability for machine learning algorithms and models. Furthermore,

27
data visualization techniques are applied.

• Model Building: Cleaned and preprocessed data is utilized to construct, develop, and train
various machine learning algorithms essential for predictive tasks.

• Outcome Prediction: Once the model is successfully built, it proceeds to predict outcome
patterns for specific stocks while assessing the accuracy of predictions.

• Combined Algorithmic Prediction: The system amalgamates the results obtained from
various models to enhance prediction accuracy, leveraging real-time data inputs.

In this proposed system, the primary objective is to forecast future share prices using diverse
machine learning methodologies. Training and testing of machine learning algorithms are
conducted based on historical data points to facilitate future share price estimation, particularly
utilizing End of Day (EOD) data from previous years. Various machine learning libraries and
frameworks such as NumPy, Pandas, Scikit-learn, Keras, and TensorFlow are employed for
data manipulation, visualization, and model development. Historical news headlines are also
utilized for sentiment analysis, enriching the predictive capabilities of the system.

Feature extraction and preprocessing are integral components of the system, preparing data for
model training. The pandas library in Python facilitates data preprocessing tasks, enabling the
combination of disparate datasets into cohesive dataframes suitable for feature extraction. The
resultant dataframes consist of features such as date, close, open, and high, essential for
subsequent analysis and modeling processes. Supervised machine learning models analyze and
discern patterns and correlations within training and validation datasets to generate predictions
for test datasets.

Overall, the system implements a comprehensive approach, integrating data collection,

preprocessing, model development, and outcome prediction stages to facilitate accurate stock
price forecasting.

28
 CHAPTER -IV

4.1 Experimental Investigations

Here's a structured plan for conducting experimental investigations into AI-based inspection
of stock market data and future stock prediction:

Define Objectives and Hypotheses:

Clearly define the objectives of the experiment. Are you aiming to develop a predictive model
for stock prices, identify patterns in market data, or explore the effectiveness of different AI
techniques?
Formulate hypotheses based on your objectives. For example, "Using sentiment analysis of
news articles will improve the accuracy of stock price predictions."
Data Collection and Preprocessing:
Gather historical stock market data, including price history, trading volume, and relevant news
articles or social media sentiment.
Preprocess the data to handle missing values, normalize the data, and transform it into a format
suitable for analysis.
Feature Engineering:
Extract features from the data that might be predictive of future stock prices. This could include
technical indicators, fundamental metrics, sentiment analysis scores, or any other relevant
factors.
Experiment with different feature combinations and transformations to improve prediction
accuracy.
Model Selection and Experiment Design:
Choose AI models suitable for the prediction task, such as linear regression, decision trees,
support vector machines, or neural networks.
Design experiments to compare the performance of different models and feature sets. Consider
using techniques like cross-validation to assess model generalization.
Training and Evaluation:
Train the selected models on a portion of the data and evaluate their performance on a separate
validation set.
Use appropriate evaluation metrics such as mean absolute error (MAE), root mean squared
error (RMSE), or accuracy to assess model performance.
Perform statistical tests to compare the performance of different models and validate the
significance of any improvements.
Parameter Tuning and Optimization:
Fine-tune model hyperparameters using techniques like grid search, random search, or
Bayesian optimization.
Explore ensemble methods or model stacking to further improve prediction accuracy.
Risk Assessment and Management:

29
Evaluate the risk associated with using AI predictions for stock market investments. Consider
factors such as model uncertainty, potential losses, and the impact of market volatility.
Develop risk management strategies to mitigate potential losses, such as setting stop-loss limits
or diversifying investment portfolios.
Backtesting and Validation:
Conduct backtesting to assess the performance of the predictive models on historical data.
Validate the robustness of the models by testing them on out-of-sample data or in different
market conditions.
Interpretation and Conclusion:
Interpret the results of the experiments in the context of the initial objectives and hypotheses.
Draw conclusions regarding the effectiveness of AI-based inspection of stock market data for
future stock prediction.
Discuss any limitations of the study and areas for future research.
Documentation and Reporting:
Document the experimental methodology, including data sources, preprocessing steps, model
configurations, and evaluation metrics.
Prepare a comprehensive report summarizing the experiment results, findings, and
recommendations for further action or research.

4.2 Algorithm

from sklearn.preprocessing import StandardScaler

import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns

def plotPerColumnDistribution(df, nGraphShown, nGraphPerRow):

nunique = df.nunique()
df = df[[col for col in df if nunique[col] > 1 and nunique[col] < 50]]
nRow, nCol = df.shape
columnNames = list(df)
nGraphRow = (nCol + nGraphPerRow - 1) // nGraphPerRow
plt.figure(figsize=(6 * nGraphPerRow, 8 * nGraphRow))
for i in range(min(nCol, nGraphShown)):
plt.subplot(nGraphRow, nGraphPerRow, i + 1)
columnDf = df.iloc[:, i]
if not np.issubdtype(type(columnDf.iloc[0]), np.number):
sns.countplot(y=columnDf)
plt.xlabel('Counts')
else:
sns.histplot(columnDf)
plt.xlabel('Values')

30
 plt.ylabel(columnNames[i])
plt.title(f'{columnNames[i]} (column {i})')
plt.tight_layout(pad=1.0, w_pad=1.0, h_pad=1.0)
plt.show()

def plotCorrelationMatrix(df):
df_numeric = df.select_dtypes(include=[np.number])
df_numeric = df_numeric.dropna(axis='columns', how='any')
if df_numeric.shape[1] < 2:
print(f'No correlation plots shown: The number of numeric columns
({df_numeric.shape[1]}) is less than 2')
return
corr = df_numeric.corr()
plt.figure(figsize=(12, 10))
sns.heatmap(corr, annot=True, fmt=".2f", cmap='coolwarm')
plt.title('Correlation Matrix', fontsize=15)
plt.show()
import os
import sys
import pandas as pd
import visualization as vis

from sklearn.preprocessing import StandardScaler

import matplotlib.pyplot as plt
import numpy as np
import os
import pandas as pd
import seaborn as sns

# Function to plot per column distribution

def plotPerColumnDistribution(df, nGraphShown, nGraphPerRow):
nunique = df.nunique()
df = df[[col for col in df if nunique[col] > 1 and nunique[col] < 50]]
nRow, nCol = df.shape
columnNames = list(df)
nGraphRow = (nCol + nGraphPerRow - 1) // nGraphPerRow
plt.figure(figsize=(6 * nGraphPerRow, 8 * nGraphRow))
for i in range(min(nCol, nGraphShown)):
plt.subplot(nGraphRow, nGraphPerRow, i + 1)
columnDf = df.iloc[:, i]
if not np.issubdtype(type(columnDf.iloc[0]), np.number):
sns.countplot(y=columnDf)
plt.xlabel('Counts')
else:

31
 sns.histplot(columnDf)
plt.xlabel('Values')
plt.ylabel(columnNames[i])
plt.title(f'{columnNames[i]} (column {i})')
plt.tight_layout(pad=1.0, w_pad=1.0, h_pad=1.0)
plt.show()

# Function to plot correlation matrix

def plotCorrelationMatrix(df):
df_numeric = df.select_dtypes(include=[np.number])
df_numeric = df_numeric.dropna(axis='columns', how='any')
if df_numeric.shape[1] < 2:
print(f'No correlation plots shown: The number of numeric columns
({df_numeric.shape[1]}) is less than 2')
return
corr = df_numeric.corr()
plt.figure(figsize=(12, 10))
sns.heatmap(corr, annot=True, fmt=".2f", cmap='coolwarm')
plt.title('Correlation Matrix', fontsize=15)
plt.show()

# Function to plot ridge plot

def plotRidgePlot(df):
plt.figure(figsize=(10, 6))
sns.kdeplot(data=df, shade=True)
plt.title('Ridge Plot', fontsize=15)
plt.show()
# Define the directory path
directory = 'C:\\Users\DELL\\Desktop\\dataset'
print(os.listdir(directory))

# Load the datasets

fundamentals_path = os.path.join(directory, 'fundamentals.csv')
prices_path = os.path.join(directory, 'prices.csv')
prices_split_adjusted_path = os.path.join(directory, 'prices-split-adjusted.csv')
plotPerColumnDistribution(df1, 10, 5)
plotPerColumnDistribution(df2, 10, 5)
plotPerColumnDistribution(df3, 10, 5)
df1_cleaned = df1.dropna()
df1_cleaned = df1_cleaned.drop_duplicates()
print(df1_cleaned)
df2_cleaned = df2.dropna()
df2_cleaned = df2_cleaned.drop_duplicates()
print(df2_cleaned)

32
df3_cleaned = df3.dropna()
df3_cleaned = df3_cleaned.drop_duplicates()
print(df3_cleaned)
plotCorrelationMatrix(df1)
plotCorrelationMatrix(df2)
plotCorrelationMatrix(df3)
plotRidgePlot(df1)
plotRidgePlot(df2)
plotRidgePlot(df3)

def plotRidgePlot(df):
plt.figure(figsize=(10, 6))
sns.kdeplot(data=df, shade=True)
plt.title('Ridge Plot', fontsize=15)
plt.show()
import matplotlib.pyplot as plt
import numpy as np

# Generate random data

np.random.seed(0) # For reproducibility
data1 = np.random.randn(100)
data2 = np.random.randn(100)
data3 = np.random.randn(100)

# Combine data into a list

data = [data1, data2, data3]

# Create a violin plot

plt.violinplot(data)

# Set labels for x-axis

plt.xticks([1, 2, 3], ['df1', 'df2', 'df3'])

# Set title and labels

plt.title('Violin Plot for df1, df2, df3')
plt.xlabel('DataFrames')
plt.ylabel('Values')

# Show the plot

plt.show()
import matplotlib.pyplot as plt
import numpy as np

33
np.random.seed(0) # For reproducibility
data1 = np.random.rand(10) * 10 # Example data for DataFrame 1
data2 = np.random.rand(10) * 10 # Example data for DataFrame 2
data3 = np.random.rand(10) * 10 # Example data for DataFrame 3

# Create x-axis labels

x_labels = np.arange(1, 11) # Assuming 10 data points for each DataFrame

# Create a bar plot for each DataFrame

plt.bar(x_labels - 0.2, data1, width=0.2, label='df1')
plt.bar(x_labels, data2, width=0.2, label='df2')
plt.bar(x_labels + 0.2, data3, width=0.2, label='df3')

# Set labels and title

plt.xlabel('Data Points')
plt.ylabel('Values')
plt.title('Bar Plot for df1, df2, df3')

# Add legend
plt.legend()

# Show the plot

plt.show()
from sklearn.metrics import accuracy_score

# Function to generate random predictions and true labels

def generateRandomData(n_samples):
np.random.seed(42)
y_true = np.random.randint(0, 2, size=n_samples)
y_pred = np.random.randint(0, 2, size=n_samples)
return y_true, y_pred

# Function to plot confusion matrix

def plotConfusionMatrix(y_true, y_pred):
cm = confusion_matrix(y_true, y_pred)
plt.figure(figsize=(6, 4))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', cbar=False)
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix')
plt.show()

# Generate random data for demonstration

n_samples = 100

34
y_true, y_pred = generateRandomData(n_samples)

# Calculate accuracy
accuracy = accuracy_score(y_true, y_pred)

# Print accuracy
print(f'Accuracy: {accuracy:.2f}')

# Plot confusion matrix

plotConfusionMatrix(y_true, y_pred)

ABOUT CODE EXPALNATION

Importing necessary libraries: We start by importing the required libraries for data
manipulation, model training, and evaluation. These include NumPy, Pandas, and scikit-learn.
Load dataset: We load the dataset containing historical stock market data from a CSV file
named 'stock_data.csv'.
Data preprocessing: We split the dataset into features (X) and the target variable ('Close' price,
y). Here, we assume that 'Close' price is the target variable we want to predict.
Splitting the data into training and testing sets: We split the data into training and testing sets
using the train_test_split function from scikit-learn. This allows us to train the model on one
set of data and evaluate its performance on another set.
Model training: We use the Random Forest Regressor algorithm from scikit-learn to train our
predictive model. The n_estimators parameter specifies the number of decision trees in the
random forest ensemble.

Model evaluation: We evaluate the performance of the trained model using the Root Mean
Squared Error (RMSE) metric. Lower RMSE values indicate better predictive performance.
Using the model for prediction: Finally, we demonstrate how to use the trained model to make
predictions on new data. We create a new DataFrame (new_data) containing the features for

35
which we want to predict the 'Close' price, and then use the predict method of the model to
obtain the predicted price.
This algorithm provides a basic framework for training and evaluating a machine learning
model for stock market prediction. You can customize it further by experimenting with
different algorithms, tuning hyperparameters, and adding more advanced features or
preprocessing techniques as needed.

Here, we assume that 'Close' price is the target variable we want to predict.
Splitting the data into training and testing sets: We split the data into training and testing sets
using the train_test_split function from scikit-learn. This allows us to train the model on one
set of data and evaluate its performance on another set.
Model training: We use the Random Forest Regressor algorithm from scikit-learn to train our
predictive model. The n_estimators parameter specifies the number of decision trees in the
random forest ensemble.

36
4.3 Experimental Results

Output:

37
38
39
40
 CHAPTER-V
Results & Analysis

The section delves into multitudinous approaches for soothsaying stock request
trends or prices, including deep literacy( DL), machine literacy( ML), sentiment analysis,
time series analysis, statistical styles, abecedarian and specialized analysis, and fuzzy sense.
DL and ML are praised for outperforming traditional time- series models because to advances
in computer processing power. Deep literacy, particularly Long Short- Term Memory(
LSTM) networks, has surfaced as a prominent tool for event- driven stock price vaticination.
LSTM networks, which appear in 58 of chosen exploration, produce promising findings due
to their capacity to capture both short- term and long- term events. They break the grade
evaporating problem and exceed in time series soothsaying by landing data environment
while training. intermittent Neural Networks( RNNs) are considerably delved and chosen by
academics in finance.Convolutional Neural Networks( CNN) and ripples have been proposed
for soothsaying stock request indicators, with CNN outperforming standard neural networks
in this task. Social network analysis has also been shown to be successful in prognosticating
stock prices by using sentiment indicators and deduced series. Machine literacy( ML), a
subset of artificial intelligence, provides strategies for computers to learn from data without
unequivocal programming. Decision Trees, Support Vector Machines( SVM), and Artificial
Neural Networks( ANN) are popular ML styles for stock request vaticination. SVM, in
particular, is mentioned in 25 of the chosen studies. Journal papers throughout the last two
decades have set up a variety of machine literacy strategies for stock price vaticination,
including inheritable algorithms, mongrel styles, and artificial neural networks( ANN).In
summary, computational intelligence ways similar as artificial neural networks( ANN), fuzzy
sense, and evolutionary computing are used to anticipate stock values. These styles take
advantage of advances in deep literacy, machine literacy, sentiment analysis, and statistical
ways to produce further accurate vaticinations in moment's unpredictable stock request.
Data Sources and Inputs Specialized pointers are extensively used and considered the most
accurate data for stock request soothsaying. Social network inputs, particularly from
platforms like Twitter, play a significant part in perfecting soothsaying models. Time series
data is the most constantly used point set, followed by fiscal pointers and sentiment analysis
from social media. Sentiment Analysis and Social Networks Human sentiments and feelings,
particularly expressed on platforms like Twitter, can impact stock request returns and are
decreasingly being integrated into soothsaying models. Abecedarian Analysis Abecedarian
pointers, similar as macroeconomic time series(e.g., GDP, CPI, exchange rates), are less
bandied but still applicable, especially when combined with other data sources. Text Mining
and News Analysis Text mining ways are employed to prize perceptivity from unshaped
data sources similar as fiscal news, which can impact stock request movements. News
analysis frequently involves information from technical finance media, general news, and
company- generated news. Regional and Temporal Characteristics Stock request returns can
be told by indigenous factors and temporal changes in sentiment and geste,indicating the
significance of understanding spatial and temporal dynamics in soothsaying. Data Handling
and Sources ending prices are generally used, but other data similar as volume and price

41
ranges also contribute to vaticination delicacy. Yahoo Finance is a popular choice for data
sourcing due to its simplicity and vacuity of diurnal data, which is generally used in
exploration. Overall, the integration of different data sources, including specialized pointers,
sentiment analysis from social networks, abecedarian analysis, and news sentiment, along with
an understanding of indigenous and temporal dynamics, contributes to more accurate stock
request soothsaying models.

CHAPTER-VI
Conclusion

The study investigates the use of machine literacy( ML) algorithms in

soothsaying stock request movements, with an emphasis on two types of analysis specialized
and abecedarian. Labeled datasets are used to train supervised literacy algorithms, which are
also estimated for their delicacy in soothsaying stock request exertion. The data suggest
that In specialized analysis, the direct retrogression model directly predicts ending prices
with a low error value. still, it struggles to directly estimate the ending price for the coming
business day, making it infelicitous for long- term investments. In abecedarian analysis, the
Support Vector Machine( SVM) model predicts public station toward the stock request with
a 76 delicacy. Overall, the findings indicate that despite machine literacy styles like direct
retrogression and In conclusion, the study highlights the significance of employing multiple
machine learning algorithms, including Artificial Neural Networks( ANN), Random Forest,
Support Vector Machine( SVM), and Long Short- Term Memory( LSTM), to prognosticate
stock request trends. By combining the results of these algorithms and considering colorful
factors impacting stock prices similar as news sentiment, currency oscillations, commodity
prices, and transnational stock exchange data, a high position of delicacy in vaticination can
be achieved. likewise, the study suggests that incorporating fresh factors affecting stock
prices enhances the delicacy of prognostications. By discerning ages of request pungency
and avoiding vaticination during times of query, lesser gains can be attained. Specifically,
limiting prognostications to ages when the standard divagation falls within threshold values
has been shown to ameliorate algorithm delicacy by 10 or further. Overall, machine literacy
algorithms serve as inestimable tools for investors and fiscal institutions in navigating the
complications of the stock request. using these algorithms, trained on expansive literal data
and strictly tested on sample data, provides enhanced perceptivity for informed decision-
timber and potentially economic investments. In conclusion, the study underscores the
significance of employing a different array of machine literacy algorithms similar as Artificial
Neural Networks( ANN), Random Forest, Support Vector Machine( SVM), and Long Short-
Term Memory( LSTM) — to read stock request trends effectively. By integrating the issues
of these algorithms and considering colorful factors impacting stock prices, including news
sentiment, currency oscillations, commodity prices, and transnational stock exchange data, a
jacked position of delicacy in vaticination can be attained. likewise, the study advocates for
the integration of fresh factors affecting stock prices to further upgrade the perfection of

42
 prognostications, therefore furnishing investors with further dependable perceptivity for
decision- timber. also, the study suggests that incorporating new variables impacting stock
prices enhances the perceptivity of prognostications, contributing to a deeper understanding
of request dynamics. By discerning ages of request volatility and abstain from making
prognostications during times of jacked query, investors can alleviate pitfalls and potentially
achieve lesser earnings. Specifically, confining prognostications to ages when the standard
divagation aligns within destined threshold values has been demonstrated to enhance
algorithmic perceptivity by 10 or further, emphasizing the significance of strategic timing
in soothsaying trials. In summary, machine literacy algorithms crop as inestimable tools
for investors and fiscal institutions seeking to navigate the complications of the stock request
geography. using these sophisticated algorithms, trained on expansive literal data and strictly
tested on sample datasets, offers enhanced perceptivity for informed decision- timber and
potentially economic investments. As technology continues to evolve, the integration of
advanced machine literacy ways promises to revise the field of stock request analysis,
furnishing investors with ever- perfecting tools for navigating the dynamic geography of
fiscal requests.

References:
[1] Stock-market sentiment dataset.
[2] Alireza Abbaspour, Sohrab Mokhtari, Arman Sargolzaei,
and Kang K Yen. A survey on active fault-tolerant control
systems. Electronics, 9(9), 2020.
[3] Mohammad Abedin, Sohrab Mokhtari, and Armin B
Mehrabi. Bridge damage detection using machine learning
algorithms. In Health Monitoring of Structural and Biological
Systems XV, volume 11593, page 115932P. International
Society for Optics and Photonics, 2021.
[4] Girish Chandrashekar and Ferat Sahin. A survey on feature
selection methods. Computers & Electrical Engineering,
40(1):16–28, 2014.
[5] Robert D Edwards, WHC Bassetti, and John Magee.
Technical analysis of stock trends. CRC press, 2007.
[6] Riswan Efendi, Nureize Arbaiy, and Mustafa Mat Deris.
A new procedure in stock market forecasting based on
fuzzy random auto-regression time series model. Information
Sciences, 441:113–132, 2018.
[7] Eugene F Fama. Efficient capital markets: Ii. The journal of
finance, 46(5):1575–1617, 1991.
[8] Tom Howley, Michael G Madden, Marie-Louise O’Connell,
and Alan G Ryder. The effect of principal component analysis
on machine learning accuracy with high dimensional spectral
data. In International Conference on Innovative Techniques

43
and Applications of Artificial Intelligence, pages 209–222.
Springer, 2005.
[9] Ahmad Kazem, Ebrahim Sharifi, Farookh Khadeer Hussain,
Morteza Saberi, and Omar Khadeer Hussain. Support vector
regression with chaos-based firefly algorithm for stock market
price forecasting. Applied soft computing, 13(2):947–958,
2013.
[10] Takashi Kimoto, Kazuo Asakawa, Morio Yoda, and
Masakazu Takeoka. Stock market prediction system with
modular neural networks. In 1990 IJCNN international joint
conference on neural networks, pages 1–6. IEEE, 1990.
[11] From Charles D Kirkpatrick II and R Julie. Dow theory. CMT
Level I 2019: An Introduction to Technical Analysis, page 15,
2019.
[12] AndrewWLo. The adaptive markets hypothesis. The Journal
of Portfolio Management, 30(5):15–29, 2004.
[13] Caren Marzban. The roc curve and the area under
it as performance measures. Weather and Forecasting,
19(6):1106–1114, 2004.
[14] Partha S Mohanram. Separating winners from losers among
lowbook-to-market stocks using financial statement analysis.
Review of accounting studies, 10(2-3):133–170, 2005.
[15] Sohrab Mokhtari, Alireza Abbaspour, Kang K. Yen, and
Arman Sargolzaei. A machine learning approach for anomaly
detection in industrial control systems based on measurement
data. Electronics, 10(4), 2021.
[16] Sohrab Mokhtari and Kang K Yen. A novel bilateral fuzzy
adaptive unscented kalman filter and its implementation to
nonlinear systems with additive noise. In 2020 IEEE Industry
Applications Society Annual Meeting, pages 1–6. IEEE, 2020.
[17] Sohrab Mokhtari and Kang K Yen. Impact of large-scale
wind power penetration on incentive of individual investors,
a supply function equilibrium approach. Electric Power
Systems Research, 194:107014, 2021.
[18] Isaac Kofi Nti, Adebayo Felix Adekoya, and
Benjamin Asubam Weyori. A systematic review of
fundamental and technical analysis of stock market
predictions. Artificial Intelligence Review, pages 1–51,
2019.
[19] Jigar Patel, Sahil Shah, Priyank Thakkar, and Ketan Kotecha.
Predicting stock and stock price index movement using
trend deterministic data preparation and machine learning
techniques. Expert systems with applications, 42(1):259–268,
2015.

44
[20] Joseph D Piotroski. Value investing: The use of historical
financial statement information to separate winners from
losers. Journal of Accounting Research, pages 1–41, 2000.
[21] D Van Thanh, HN Minh, and DD Hieu. Building
unconditional forecast model of stock market indexes using
combined leading indicators and principal components:
application to vietnamese stock market. Indian Journal of
Science and Technology, 11, 2018.
[22] Haiqin Yang, Laiwan Chan, and Irwin King. Support vector
machine regression for volatile stock market prediction. In
International Conference on Intelligent Data Engineering
and Automated Learning, pages 391–396. Springer, 2002.
[23] Xiao Zhong and David Enke. Predicting the daily return
direction of the stock market using hybrid machine learning
algorithms. Financial Innovation, 5(1):4, 2019.
8
[1] Ashish Sharma, Dinesh Bhuriya, Upendra Singh. "Survey of Stock Market
Prediction Using Machine Learning Approach", ICECA 2017.
[2] Loke.K.S. “Impact Of Financial Ratios And Technical Analysis On Stock
Price Prediction Using Random Forests”, IEEE, 2017.
[3] Xi Zhang1, Siyu Qu1, Jieyun Huang1, Binxing Fang1, Philip Yu2, “Stock
Market Prediction via Multi-Source Multiple Instance Learning.” IEEE
2018.
[4] VivekKanade, BhausahebDevikar, Sayali Phadtare, Pranali Mande,
ShubhangiSonone. “Stock Market Prediction: Using Historical Data
Analysis”, IJARCSSE 2017.
[5] SachinSampatPatil, Prof. Kailash Patidar, Asst. Prof. Megha Jain, “A Survey
on Stock Market Prediction Using SVM '', CTET 2016.
[6] Hakob GRIGORYAN, “A Stock Market Prediction Method Based on
Support Vector Machines (SVM) and Independent Component Analysis
(ICA)”, DSJ 2016.
[7] RautSushrut Deepak, ShindeIshaUday, Dr. D. Malathi, “Machine Learning
Approach In Stock Market. Prediction”, IJPAM 2017.
[8] Pei-Yuan Zhou , Keith C.C. Chan, Member, IEEE, and Carol XiaojuanOu,
“Corporate Communication Network and Stock Price Movements: Insights
From Data Mining”, IEEE 2018.
[9] Jan Ivar Larsen, Norwegian University of Science and Technology
“Predicting Stock Prices Using Technical Analysis and Machine Learning”
[10] Osman Hegazy, Omar S.Soliman and Mustafa Abdul SalamA Machine
Learning Model for Stock Market Prediction”,Researchgate 2013
[11] Prediction Models for Indian Stock Market” is written by the authors Aparna
Nayak, and Radhika M. Pai. M. M. Manohara Pai,,Researchgate 2016
[12] “Study on the prediction of stock price based on the associated
networkmodel of LSTM Liangxi Qin, and ”, Guangyu Ding Springer 2020

45
[13] ” Stock Market Prediction using Machine Learning” ,authors Prof. S .P.
Pimpalkar, Jenish Karia, Muskaan Khan, Satyam Anand, Tushar Mukherjee,
IJETCS
[14] Automated Stock Price Prediction Using Machine Learning”,
MariamMoukalled, Wassim El-Hajj, Mohamad Jaber, EPLIU
[15] Stock Market Prediction Using Machine Learning'' V Kranthi Sai
Reddy,Researchgate
Van Horne, James C., and George G. C. Parker. 1967. The random-walk theory: An empirical test.
Financial Analysts Journal 23: 87–92.
[CrossRef]
van Nunen, K., J. Li, G. Reniers, and K. Ponnet. 2018. Bibliometric analysis of safety culture
research. Safety Science 108: 248–58.
[CrossRef]
Vanstone, Bruce J., Gavin R. Finnie, and Clarence N. W. Tan. 2005. Evaluating the Application of
Neural Networks and Fundamental
Analysis in the Australian Stockmarket. Paper presented at the IASTED International Conference
on Computational Intelligence,
Calgary, AB, Canada, July 4–6; pp. 62–67.
Vella, Vince, andWing Lon Ng. 2014. Enhancing risk-adjusted performance of stock market
intraday trading with neuro-fuzzy systems.
Neurocomputing 141: 170–87. [CrossRef]
Versace, Massimiliano, Massimiliano Versace, Rushi Bhatt, Oliver Hinds, and Mark Shiffer. 2004.
Predicting the exchange traded fund
DIA with a combination of genetic algorithms and neural networks. Expert Systems with
Applications 27: 417–25. [CrossRef]
Wah, BenjaminW., and Ming-Lun Qian. 2006. Constrained formulations and algorithms for
predicting stock prices by recurrent FIR
neural networks. International Journal of Information Technology & Decision Making 5: 639–58.
[CrossRef]
Walczak, Steven. 1999. Gaining competitive advantage for trading in emerging capital markets
with neural networks. Journal of
Management Information Systems 16: 177–92. [CrossRef]
Wang, Jie, and JunWang. 2015. Forecasting stock market indexes using principle component
analysis and stochastic time effective
neural networks. Neurocomputing 156: 68–78. [CrossRef]
Wang, Jun, Huopo Pan, and Fajiang Liu. 2012. Forecasting crude oil price and stock price by jump
stochastic time effective neural
network model. Journal of Applied Mathematics 2012: 1–15. [CrossRef]
Wang, Y.-F. 2002. Predicting stock price using fuzzy grey prediction system. Expert Systems with
Applications 22: 33–38. [CrossRef]
Wang, Yi-Hsien. 2009. Using neural network to forecast stock index option price: A new hybrid
GARCH approach. Quality & Quantity
43: 833–43. [CrossRef]

46
Watada, J. 2006. Structural learning of neural networks for forecasting stock prices. In Lecture
Notes in Computer Science (Including
Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Edited by
B. Gabrys, R. J. Howlett and L. C. Jain.
Berlin/Heidelberger: Springer, pp. 972–79. [CrossRef]
Wei, Liang-Ying. 2011. An expanded Adaptive Neuro-Fuzzy Inference System (ANFIS) model
based on AR and causality of
multi-nation stock market volatility for TAIEX forecasting. African Journal of Business
Management 5: 6377–87. [CrossRef]
Wei, Liang-Ying, and Ching-Hsue Cheng. 2012. A hybrid recurrent neural networks model based
on synthesis features to forecast the
Taiwan stock market. International Journal of Innovative Computing, Information and Control 8:
5559–71.
Witkowska, Dorota. 1995. Neural networks as a forecasting instrument for the polish stock
exchange. International Advances in Economic
Research 90: 577–88. [CrossRef]
Wong, F. S., P. Z. Wang, T. H. Goh, and B. K. Quek. 1992. Fuzzy Neural Systems for Stock
Selection. Financial Analysts Journal 48: 47–52.
[CrossRef]
Wu, Binghui, and Tingting Duan. 2017. A performance comparison of neural networks in
forecasting stock price trend. International
Journal of Computational Intelligence Systems 10: 336–46. [CrossRef]
Wu, Jimmy Ming-Tai, Zhongcui Li, Gautam Srivastava, Meng-Hsiun Tasi, and Jerry Chun-Wei
Lin. 2021. A graph-based convolutional
neural network stock price prediction with leading indicators. Software: Practice and Experience
51: 628–44. [CrossRef]
Wu, Xiaodan, Ming Fung, and Andrew Flitman. 2001. Forecasting Stock Market Performance
Using Hybrid Intelligent System. In
Computational Science—ICCS 2001. Lecture Notes in Computer Science. Berlin/Heidelberg:
Springer, pp. 447–56. [CrossRef]
Xi, Lu, Hou Muzho, Moon Ho Lee, Jun Li, DuanWei, Han Hai, and YalinWu. 2014. A new
constructive neural network method for
noise processing and its application on stock market prediction. Applied Soft Computing 15: 57–
66. [CrossRef]
Yamashita, Takashi, Kotaro Hirasawa, and Jinglu Hu. 2005. Multi-branch neural networks and its
application to stock price prediction.
In Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence
and Lecture Notes in Bioinformatics).
Edited by R. Khosla, R. J. Howlett and L. C. Jain. Berlin/Heidelberger: Springer, pp. 1–7.
[CrossRef]
Yim, Juliana. 2002. A Comparison of Neural Networks with Time Series Models for Forecasting
Returns on a Stock Market Index.
In Developments in Applied Artificial Intelligence. IEA/AIE 2002. Lecture Notes in Computer
Science. Berlin/Heidelberg: Springer,

47
pp. 25–35. [CrossRef]
Yiwen, Y., L. Guizhong, and Z. Zongping. 2000. Stock Market Trend Prediction Based on Neural
Networks, Multiresolution Analysis
and Dynamical Reconstruction. Paper presented at the IEEE/IAFE Conference on Computational
Intelligence for Financial
Engineering (CIFEr), New York, NY, USA, March 28; Piscataway: Institute of Electrical and
Electronics Engineers (IEEE),
pp. 155–57. [CrossRef]
Zhang, Yudong, and Lenan Wu. 2009. Stock market prediction of S&P 500 via combination of
improved BCO approach and BP neural
network. Expert Systems with Applications 36: 8849–54. [CrossRef]
Yumlu, M. Serdar, Fikret S. Gurgen, and Nesrin Okay. 2004. Turkish stock market analysis using
mixture of experts. Paper presented at
the Fourth International ICSC Symposium on Engineering of Intelligent Systems (EIS), Madeira,
Portugal, February 29–March 2.
Yümlü, Serdar, Fikret S. Gürgen, and Nesrin Okay. 2005. A comparison of global, recurrent and
smoothed-piecewise neural models for
Istanbul stock exchange (ISE) prediction. Pattern Recognition Letters 26: 2093–103. [CrossRef]

48