International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

ISSN (online): 2347-5552, Volume-11, Issue-6, November 2023

https://doi.org/10.55524/ijircst.2023.11.6.4
Article ID IRP1414, Pages 18-25
www.ijircst.org

Applications of Machine Learning in Predictive Analysis and

Risk Management in Trading
Kavin Karthik V
B.Tech Scholar, School of Computer Science and Engineering (SCOPE), Vellore Institute of Technology,
Chennai, Tamil Nadu, India
Correspondence should be addressed to Kavin Karthik V;
Received 23 October 2023; Revised 7 November 2023; Accepted 20 November 2023
Copyright © 2023 Kavin Karthik V . This is an open-access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

ABSTRACT- The stock market is considered the historical progression and unveiling the profound
primary domain of importance in the financial sector implications for modern trading strategies. The paper is
where Artificial Intelligence combined with various structured to provide a comprehensive understanding of
algorithmic practices empowers investors with data- two pivotal facets: risk management in equity markets
driven insights, enhancing decision-making, predicting and predictive analysis of stock trends through the
trends, and optimizing risk management for more application of cutting-edge machine learning models,
informed and strategic financial outcomes. This research including Convolutional Neural Networks (CNN), Long
paper delves into the real-world applications of machine Short-Term Memory networks (LSTM), and Deep
learning and algorithmic trading, observing their Reinforcement Learning (DRL).
historical evolution together and how both of these can go The historical roots of AI in trading trace back to the mid-
hand in hand to control risk and forecast the movement of 20th century when early AI pioneers, like Marvin Minsky
a stock or an index and its future. The research is and John McCarthy, envisaged the potential of machines
structured to provide comprehensive insights into two to replicate human decision-making in complex financial
major subdomains in the application of AI in algorithmic domains. In recent decades, advancements in
trading: risk management in equity markets and computational power, access to extensive historical
predictive analysis of stock trends through the application financial data, and the development of sophisticated
of machine learning models and training the current machine learning algorithms have propelled the
existing data which is feasible and training them with integration of AI in trading systems. This evolution has
respect to historical scenarios of various market trends given rise to algorithmic trading, high-frequency trading,
along with various fundamental and technical analysis and quantitative strategies, altering the landscape of
techniques with the help of various deep learning financial markets [1].
algorithms. For risk management of a portfolio in finance, Risk management and portfolio optimization have
various machine learning models can be employed, experienced a transformative historical journey.
depending on the specific needs and goals of the portfolio Traditionally, risk management relied on statistical
manager or risk analyst and implementing various value- methods and historical data to estimate portfolio risk.
at-risk algorithms along with deep learning techniques in Markowitz's Mean-Variance Optimization (MVO),
order to assess risk at particular trade position and to introduced in 1952, revolutionized portfolio optimization.
manage volatile trades at unprecedented situations. The MVO aimed to strike an optimal balance between risk
significance of this research paper lies in its practical and return by considering asset correlations. However, it
applicability, offering real-world solutions to enhance had limitations in handling non-linear relationships and
trading strategies and decision-making processes with a outliers, which are prevalent in financial markets. The
focus on mitigating risk and capitalizing on market historical development of AI techniques in portfolio
opportunities and also giving clear insights with respect optimization can be traced back to the late 20th
to the current practical limitations of application of the century[4]. Neural networks emerged as promising tools
provided solution and future scope to overcome the same. for modeling complex financial data. Multilayer
perceptrons, a type of neural network, could capture non-
KEYWORDS- Algorithmic Trading, Risk linear patterns and improve risk estimation. Nevertheless,
Management, Equity Markets, Portfolio Management, their training process required extensive data, making
Predictive Analysis, Fundamental Analysis, Value at them less suitable for small datasets. Convolutional
Risk. Neural Networks (CNNs), initially designed for image
analysis, found their way into financial markets. CNNs
I. INTRODUCTION can process time series data and identify complex
patterns, which is pivotal for portfolio optimization.[9]
The intersection of artificial intelligence (AI) and
For example, they can analyze historical price charts and
financial trading has evolved into an intricate symbiosis,
uncover non-linear relationships between assets,
catalyzing the transformation of traditional trading
enhancing risk estimation.
practices. This research paper embarks on an in-depth
The way of finding the future valuation of the stock
exploration of this dynamic relationship, dissecting its
market prices is called the stock market estimate.

Innovative Research Publication 18

 International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

Expected to be Strong, accurate, and effective. The develop adaptive trading strategies that can capitalize on
system should work in line with real-life scenarios and be changing market conditions. This adaptability makes
well connected to that. The movement in the stock market DRL a valuable tool for traders seeking to navigate the
is usually determined by the sentiments of thousands of complexities of financial markets and make decisions in
investors [2]. These events are political events such as the real time based on the evolving market landscape.
statements of ministers or government officials,
statements of government bodies such as RBI, SEBI, II. LITERATURE SURVEY
scandal news, etc. It can also be the global happening
such as rapid movements in currencies prices and The paper is divided into 5 major sections: Historical
commodities prices [3]. All this thing affects the earnings evolution and the current trend of AI and Algorithmic
of companies, which ultimately affects the sentiment of Trading, Previous research and articles in the same
stock market investors. This is beyond the reach of almost domain, Methodology of assessing risk in executing a
all individuals to assess these accurately and consistently. trade, Methodology to forecast a trade, Real-world
This method usually requires the collection of various application of the methodology and algorithm,
social media data, news that affects stock market Limitations in the methodologies and Scope for future
investors sentiment, and the feelings expressed by improvement. My contribution to this domain of research
individuals. Other data such as last year’s stock prices are majorly involves Long Short-Term Memory (LSTM) and
also considered. The relationship between different data deep learning algorithms in the context of stock market
points. Is considered and an estimation is done using predictive analysis and making use of Collaborative
these variety of data points. Robust risk management in Filtering and value-at-risk methodologies for risk
equity markets is a cornerstone of successful trading. AI management in portfolio optimization.
contributes significantly by enabling the development of Nayak et al [17] highlight precision and accuracy in
advanced risk models, including Value at Risk (VaR) and predicting share prices. While various methods, including
Conditional Value at Risk (CVaR) [2]. These models time series analysis, fundamental analysis, and technical
employ machine learning and statistical techniques to analysis, have been employed by investors and
assess market volatility, optimize portfolio institutions, they are not always reliable. To address this
diversification, and formulate effective hedging strategies challenge, the paper introduces a Machine Learning (ML)
[4]. Predictive analysis of stock trends is a fundamental approach, which is trained on historical stock data to
concern for traders and investors. CNN, a neural network acquire intelligence and make accurate predictions. After
architecture designed for image processing, can be extensive research, the paper finds that the Artificial
adapted to identify patterns in financial time series data. Neural Network (ANN) is more suitable than other
[5] LSTM, a specialized recurrent neural network, excels algorithms. A customized neural network model is
in modeling sequential data and is well-suited for employed, and the method is tested on the Bombay Stock
predicting stock price movements. DRL, inspired by Exchange index dataset
reinforcement learning, offers a dynamic approach to Ding, G., et al. [18] explores a two-layer approach
trading by learning optimal strategies through interactions incorporating technical analysis and machine learning.
with the market environment [6]. The model described in the paper incorporates a two-
Convolutional Neural Networks (CNN) in Stock Price layer approach, with the first layer relying on technical
Prediction: CNN, originally developed for image analysis and the second on machine learning.
processing, has found a unique application in financial Additionally, the paper introduces a financial
markets. These networks can analyze historical price management strategy that considers the historical success
charts, identifying patterns and trends that may not be of predictions to determine future investments. Through
apparent to human analysts. By processing the pixel-level portfolio simulations and trading models, the research
data of price charts, CNN models can extract valuable concludes that the predictive model effectively surpasses
information to predict short-term and long-term stock the Oslo Benchmark Index (OSEBX) [7]
price movements. Their ability to recognize complex Birant, D et al [19] proposes a model combining particle
patterns in financial time series data makes them a swarm optimization and random forest algorithms for
valuable tool for traders and investors [7]. Long Short- daily share market price prediction.[9] This model works
Term Memory Networks (LSTM) in Time Series with historical share market data and technical indicators.
Analysis: LSTM, a specialized recurrent neural network By avoiding local minima and improving prediction
architecture, has gained immense popularity in time series accuracy, the particle swarm optimization algorithm plays
forecasting. These networks are well-suited for modeling a key role. The model is tested on multiple financial
sequential data, which is a key characteristic of financial datasets and compared with Leuralberg-Marquardt neural
time series. LSTM models can capture dependencies over network algorithms, resulting in improved prediction
time, making them effective for predicting stock prices accuracy.
and trends. Their ability to remember and learn from past A. Nayak, M. M. M. Pai, and R. M. Pai's "Prediction
data enables them to adapt to changing market conditions Models for Indian Stock Market" [17] discusses
and provide accurate forecasts [8]. Deep Reinforcement predictive models for next-day and one-month share price
Learning (DRL) for Adaptive Trading: Deep predictions using supervised machine learning. Daily
Reinforcement Learning, inspired by reinforcement predictive models rely on historical data, and machine
learning principles, offers a dynamic approach to trading. learning techniques yield up to 70% accuracy. Monthly
DRL agents learn optimal strategies by interacting with forecasting models attempt to identify patterns between
financial markets, receiving rewards for profitable actions different months. Tests indicate that the trend for at least
and penalties for losses[9]. Over time, these agents one month is correlated with other months' trends,

Innovative Research Publication 19

 International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

reinforcing the importance of machine learning in ensemble classifier-based recommender systems surpass
predicting share market prices. single classifiers in predicting stock price movements.
G. Ding and L. Qin's "Study on the Prediction of Stock Prieto-Torres and Galpin [24] used a CF model to build a
Price Based on Associated Network Model of LSTM" virtual wallet recommender system for a FinTech
[18] explores multiple methods to estimate future share company to enhance the user experience on their mobile
prices, including various AI techniques. The paper application. However, this method worked best using a
introduces a deep neural network model based on Long non-personalized approach where the top-N most popular
Short-Term Memory (LSTM) networks with multiple stocks were recommended to each user. Anwar, et.al,
inputs and outputs. This model can simultaneously (2019) proposed a collaborative filtering-based movie
estimate open, high, and low share prices. The results recommendation using rule mining to satisfy the
demonstrate that the associated model surpasses other individual requirement. Their paper also provided a
models in accuracy, achieving an estimated accuracy of comparative analysis of the several types of similarity
over 95%. measures namely Co- sine, Correlation, Euclidean,
S. P. Pimpalkar, Jenish Karia, Muskaan Khan, Jaccard, and Manhattan. The correlation similarity
Satyamandand, and Tushar Mukherjee's "Stock Market measure yielded the best precision, recall, and F1-score.
Forecasts Using Machine Learning" [22] explores the Further, Anwar, et.al, (2021) introduced a new approach
construction of predictive models using various attributes, that uses CF and Singular Value Decomposition (SVD)
including oil prices, foreign exchange rates, interest rates, ++ for implementing a recommendation system. Their
gold and silver prices, news, Twitter news feeds, and proposed approach gave a smaller error rate when cross-
pattern matching. The paper employs different Machine validation (CV = {5, 10, 15}) was performed. Their
Learning (ML) methods, such as Support Vector proposed approach also alleviated the sparsity and cold-
Machines and Recurrent Neural Networks, to predict start problems.
market movements based on these attributes. Kirubahari, R, et.al [23] proposed a weighted parallel
M. Moukalled, W. El-Hajj, and M. Jaber's "Automated deep hybrid collaborative filtering approach based on
Stock Price Prediction Using Machine Learning" [20] Singular Value Decomposition (SVD) and Restricted
highlight the importance of news in share market Boltzmann Machine (RBM). Their results indicated better
predictions. The authors employ conventional machine prediction compared to other techniques in terms of ac-
learning models and deep learning models to predict curacy. Cui et. al (2019) also explored the traditional CF
stock prices by considering relevant issues. The research approach as well as SVD to ease fund companies’
reveals that a Support Vector Machine model achieves decision-making process in identifying which listed
the highest accuracy of 82.91% for predicting stock company to invest in. In their study, they tackled the issue
prices. of sparsity for user-based CF by taking into account the
V. V. K. Sai Reddy's "Stock Market Forecasts Using number of items that are similarly scored by the users and
Machine Learning" [21] emphasizes the importance of improved traditional item-based CF by inputting the
stock market prediction, employing Machine Learning missing ratings of the users through an iterative process
(ML) methods to predict major and minor stock prices in instead of just using the mean ratings. Their study showed
different markets. The paper introduces a Support Vector that the improved item-based CF outperforms the other
Machine for predicting share prices using daily and two methods.
minute wave prices [16]. Lastly, Rukiya et. al [25] proposes an improvement to the
When it comes to Risk Management, Collaborative traditional CF approach by first clustering based on the
filtering (CF) is one of the most popular techniques used popularity of the items and segmenting the users based on
to build recommender systems. It works on the their loyalty to the business to provide a more
assumption that users with similar preferences in the past personalized recommendation while ensuring that items
are more likely to have similar interests in the future. that are already known to the user are not recommended.
There are two categories: User-based and Item-based. This allows e-commerce platforms’ targeted marketing
User-based CF finds users with similar consumption approaches to be more effective, garnering greater user
patterns and recommends items that these similar users loyalty. Content-based filtering (CBF) is another
find interesting. User-based CF performance decreases technique used to build recommender systems. It uses the
when we have sparse datasets ( Boström & Filipsson, information about the items obtained from the user’s
2017 ). On the other hand, item-based CF recommends profile to recommend similar items to what the user has
items similar to what the user has liked in the past. Since liked in the past. [12] Unlike CF, CBF does not take into
the properties of items remain more constant compared to account other users’ information. Yoo et al. (2003)
users’ evolving tastes and preferences, item-based CF proposed a personalized filtering system that only
tends to be more scalable than user-based CF ( Boström considers the trading information that a user deems
& Filipsson, 2017 ). relevant. Personalized recommendations based on the
Sayyed et al. (2013) [22] presented a preliminary Moving Average Convergence Divergence (MACD)
investigation on the implementation of CF methods on advise traders whether to buy or sell a stock by examining
the stock market but had not validated CF methods on the difference between two moving averages. Chalida
stock data. Vismayaa et al. (2020) compared the bhongse and Kaensar (2006) proposed an adaptive user
performance of single classifiers against ensemble model that performs stochastic technical analysis to
classifier-based recommender systems on stocks from the forecast stock returns while considering explicit
Bombay Stock Exchange based on classification accuracy preferences and user interactions with the system for
and economic measures. This study showed that personalized recommendations. Rutkowski (2021) built a
Neuro-Fuzzy recommender system to recommend a stock

Innovative Research Publication 20

 International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

based on similarity from a user’s past transaction if the challenge faced for the processing of the textual data are
stock passed the fuzzy set of rules generated. that the information generated on these platforms is
enormous, increasing the computational complexities. For
III. METHODOLOGY example, the researchers in [13] processed 1,00,000
tweets, and the researchers in [12] processed around
SMP systems can be classified according to the type of 2,500,000 tweets, which was a complex task. Moreover,
data they use as the input. Most of the studies used for the textual data, no proper standard format is followed
market data for their analysis. Recent studies have while posting on social media, which increases the
considered textual data from online sources as well. In processing complexities. In addition, the detection of
this section, the studies are classified based on the type of shorthand spellings, emoticons, and sarcastic statements
data they use for prediction purposes. are yet another challenge. Machine learning algorithms
Market Data Market data are the temporal historical come to the forefront to deal with all kinds of challenges
price-related numerical data of financial markets. faced while processing textual data.
Analysts and traders use the data to analyze the historical
trend and the latest stock prices in the market. They
reflect the information needed for the understanding of
market behavior. The market data are usually free, and
can be directly downloaded from the market websites.
Various researchers have used this data for the prediction
of price movements using machine learning algorithms.
Previous studies have focused on two types of
predictions. Some studies have used stock index
predictions like the Dow Jones Industrial Average (DJIA)
[7], Nifty [9], Standard and Poor’s (S&P) 500 [10],
National Association of Securities Dealers Automated
Quotations (NASDAQ) [12], the Deutscher Aktien Index
(DAX) index [13], and multiple indices [14,15]. Other
studies have used individual stock prediction based on
some specific companies like Apple, Google, or groups of
companies. Furthermore, the studies focused on time-
specific predictions like intraday, daily, weekly, and
monthly predictions, and so on. Moreover, most of the
previous research is based on categorical prediction,
where predictions are categorized into discrete classes
like up, down, positive, or negative [13]. Technical Figure 1: Generic Scheme for Stock Market Prediction
indicators have been widely used for SMP due to their using ML Algorithms
summative representation of trends in time series data.
Some studies considered different types of technical
indicators, e.g., trend indicators, momentum indicators, IV. DATA PREPROCESSING
volatility indicators, and volume indicators. Furthermore,
Once the data is available, it needs some pre-processing
numerous studies have used an amalgam of different
so that it can be fed to a machine-learning model. The
types of technical indicators for SMP.
significance of the output depends on the pre-processing
Textual Data Textual data is used to analyze the effect of
of the data [15]. The textual data must be transformed
sentiments on the stock market. Public sentiments have
into a structured format that can be used in a machine-
been proven to affect the market considerably. The most
learning model. The previous studies revealed that there
challenging part is to convert the textual information into
are three significant pre-processing steps, i.e., feature
numerical values so that it can be fed to a prediction
selection, order reduction, and the representation of
model. Furthermore, the extraction of textual data is a
features. Table 1 presents the comparison of the data
challenging task. The textual data has many sources, such
sources, type of input, and prediction duration. Table 2
as financial news websites, general news, and social
presents the comparison of the data pre-processing
platforms. Most of the studies were carried out on textual
techniques used in the studies so far and how each of the
data to try to predict whether the sentiment towards a
techniques contributes to enhancing accuracy in each
particular stock is positive or negative. The previous
method.
studies considered several textual sources for SMP, such
as the Wall Street Journal, Bloomberg, CNBC and
Reuters, Google Finance, and Yahoo Finance. The
extracted news may be either generalized news or some
specific financial news, but the majority of the
researchers use financial news, as it is deemed to be less
susceptible to noise. Some researchers have used less
formal textual data, such as message boards. Meanwhile,
the textual data from microblogging websites and social
networking websites are comparably less explored than
other textual data forms for SMP. Besides this, one

Innovative Research Publication 21

 International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

Table 1: Comparison of the data sources, type of input, and prediction

Table 2: Comparison of the data pre-processing techniques

contributes to a deeper understanding of market sentiment

V. PROPOSED SYSTEM and enhances the prediction accuracy.
In the proposed system, our primary objective is to
predict future share prices using various advanced
machine learning methods, with a specific focus on deep In our data processing pipeline, the NumPy library played
neural networks like LSTM (Long Short-Term Memory). a pivotal role in preparing and cleaning the datasets. It
Our approach involves training and testing machine allowed us
learning algorithms on historical data points, allowing us to transform the data into a format that can be directly
to estimate future share prices with accuracy and used by the machine learning models [6].
precision [15]. To accomplish this, we leveraged a rich The Sklearn library is essential for actual calculations,
dataset comprising EOD (End of the Day) historical data estimations, and predictions. It serves as a reliable tool
spanning several years. This dataset serves as the for performing various machine-learning tasks, ensuring
foundation for training and testing our machine-learning that our predictive models are robust and accurate [4]. We
models [15]. We employed a range of machine learning sourced historical share market data from various public
libraries and frameworks to achieve our goals, including and open online repositories, using approximately 85% of
NumPy and Pandas for data manipulation and the dataset for training the machine learning models, with
visualization, Scikit-Learn (Sklearn) and TensorFlow for the remainder allocated for validation and testing
building machine learning models, and the math library purposes [4].
for performing mathematical operations on the data [6]. The core approach of our supervised machine learning
A significant component of our dataset for sentiment models is to analyze and recognize patterns and
analysis was historical news headlines. This data source correlations within the dataset. By training on both the

Innovative Research Publication 22

 International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

training and validation datasets, the models learn to Table 3: Results for the R-Squared value (R2)
produce accurate predictions for the test dataset [13].Our
system undertakes feature extraction and various data
pre-processing steps to prepare the data for model
training. The Python library, Pandas, played a vital role in
combining various datasets into a single, comprehensive
data frame. This pre-processed data frame includes
critical features such as date, close, open, high, low,
volume, delivery percentage, number of trades, turnover,
and other derived features. All these features are used to
train the machine learning models, with a particular focus
on a random forest model. Additionally, we utilize
selected features to make predictions using advanced
algorithms like LSTM and SVM for forecasting the share
prices of the upcoming day [14]. The accuracy of these
predictions is thoroughly assessed on the test dataset and
compared against real values. Furthermore, our system
incorporates Artificial Neural Networks (ANNs) for Table 3 displays the tabulated results for the R-Squared
sentiment analysis of news data. We employ the Google value (R2) acquired when a particular model was tested
News API to fetch real-time news data related to the for that particular company’s dataset. With the ideal value
stock market. The analysis of this news data is crucial in for R-squared being close to a non-negative ‘1’, from this
understanding the market sentiment, which is a key factor table, it is observed that out of all the five different
in predicting share prices. In terms of risk management, algorithmic models, the DL algorithm i.e., the Long
our system integrates risk management algorithms to Short-Term Memory algorithm has provided the best
ensure that investments are well-balanced and aligned results, as it has R-squared quite close 1, (i.e., -0.11),
with the risk tolerance of the investor. These algorithms followed by Support Vector Regression, with an R-
analyze portfolio composition and dynamically adjust it squared value of -1.69, so on and so forth.
based on real-time market conditions, further enhancing
the robustness of our system. Table 4: Results for the SMAPE acquired
Overall, our proposed system encompasses various
technical aspects, including data manipulation, feature
extraction, sentiment analysis using ANNs, advanced
machine learning techniques like LSTM, SVM, and risk
management algorithms, all aimed at providing accurate
and reliable predictions for share prices in the dynamic
stock market environment.

VI. RESULTS AND DISCUSSION

In this section, the results after carrying out the successful
implementation of the project have been presented in the
form of tables and graphs. As part of this project, five
algorithms i.e., the K-Nearest Neighbour algorithm,
Support Vector Regression algorithm, Linear Regression
algorithm, Decision Tree Regression algorithm, and Long
short-term memory algorithm were chosen for the
prediction of stock prices of twelve different companies.
The dataset was huge, starting from 2015 up to and Table 4 displays the tabulated results for the SMAPE
including 2021. The models were tested for 8 trading acquired when a particular model was tested for that
days i.e., since the train-to-test ratio was 99:1, and the particular company’s dataset. With the ideal value for
data for 2304 days was used for training whereas the SMAPE being close to zero, from this table, it is observed
remaining 8 days were allocated for testing the created that out of all the five different algorithmic models, the
models. The models were tested on three essential DL algorithm i.e., the Long Short-Term Memory
performance metrics, namely, Symmetric Mean Absolute algorithm has rendered the best predictive performance,
Percentage Error (SMAPE), R-squared value (R2), and as it has the least value of error (1.59), followed by
Root Mean Square Error (RMSE). These are well-known Support Vector Regression, with a SMAPE of 5.59 etc.
prevalent evaluation parameters that help researchers to
draw conclusions about the different models that were VII. FUTURE SCOPE AND LIMITATIONS
being studied. OF PROPOSED MODEL
With the increasing demand for ML in almost every
possible place and situation, be it industries business
models, or healthcare domains, it is of utmost importance
to make better models that can make more accurate and
precise predictions from huge sets of data available.

Innovative Research Publication 23

 International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

However, the analysis from the results above and various Short-Term Memory algorithm was a DL algorithm that
other literature reviews suggests that ML yields less provided the best results during stock price prediction.
authentic results when it comes to the prediction of time
series data. Nevertheless, there is a potent solution to this REFERENCES
which lies in DL and neural networks that offer great
1. Russell, S. J., & Norvig, P. (2016). Artificial Intelligence:
results during time series prediction when compared to
A Modern Approach (3rd ed.). Pearson.
ML. The results found after the implementation of this 2. Chan, E. P. (2013). Algorithmic Trading: Winning
project conform to the theoretical facts related to the Strategies and Their Rationale. John Wiley & Sons.
performance of the algorithms. As far as this project is 3. Ashish Sharma, Dinesh Bhuriya, Upendra Singh. "Survey
concerned, there may be more algorithms implemented in of Stock Market Prediction Using Machine Learning
the future and different datasets may be used in order to Approach", ICECA 2017.
offer a wider spectrum for comparison of the algorithms 4. McNeil, A. J., Frey, R., & Embrechts, P. (2015).
and on a wider note comparison of ML and DL. DL Quantitative Risk Management: Concepts, Techniques,
techniques comprise several layers which drive them a and Tools (Revised ed.). Princeton University Press.
5. Zhang, Y., Zhao, D., & Dong, F. (2018). Stock Price
step closer to the way human brains function whereas, on
Prediction with Convolutional Neural Network. In
the contrary, ML still requires a lot more human Proceedings of the International Conference on Neural
assistance. DL techniques are based on the concept of Information Processing (ICONIP) (pp. 843-850).
neural networks, similar to the neuron cells in the human 6. Hochreiter, S., & Schmidhuber, J. (1997). Long Short-
brain. Further DL techniques equip the machine to Term Memory. Neural Computation, 9(8), 1735-1780.
identify features that make identification easier or in other 7. Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A.,
words hierarchical arrangement of the features. A major Antonoglou, I., Wierstra, D., & Riedmiller, M. (2015).
advantage that comes with DL is that they continue to Human-level Control through Deep Reinforcement
become more robust and efficient with increasing data, Learning. Nature, 518(7540), 529-533.
8. Zhang, Y., Zhao, D., & Dong, F. (2018). Stock Price
which is a quite favorable feature for data-extensive
Prediction with Convolutional Neural Network. In
projects and historic time-series predictions. The future Proceedings of the International Conference on Neural
lies in the development and betterment of more and more Information Processing (ICONIP) (pp. 843-850).
ML and DL-based models requiring minimum human 9. Park, S., Kihara, K., & Watanabe, K. (2015). Portfolio
intervention, lesser prediction time, more accurate Risk Assessment Using Support Vector Machines. In
predictions, the capability to handle huge data with the Proceedings of the International Conference on Neural
help of multiple layers, reduced complexity, and more Information Processing (ICONIP) (pp. 226-233).
affordability, etc. The upcoming researchers possess DL 10. Bansal, M., Goyal, A., & Choudhary, A. (2022). Stock
and ML possess high potential for the upcoming market prediction with high accuracy using machine
learning techniques. Procedia Computer Science, 215,
researchers to study, scientifically curate, and develop
247-265.
much more efficient and robust models that could predict 11. Rouf, N., Malik, M. B., Arif, T., Sharma, S., Singh, S.,
faster and more precise results based on real-time Aich, S., & Kim, H. C. (2021). Stock market prediction
situations and would help the world save and progress in using machine learning techniques: a decade survey on
multiple domains to give rise to holistic development in methodologies, recent developments, and future
interdisciplinary areas with minimum to no human directions. Electronics, 10(21), 2717.
intervention 12. Chandrinos, S. K., Sakkas, G., & Lagaros, N. D. (2018).
This research work aimed at developing ML models that AIRMS: A risk management tool using machine
would be capable of predicting stock prices with learning. Expert Systems with Applications, 105, 34-48.
13. Grote, M., & Bogner, J. (2023). A Case Study on AI
increased accuracy so that interested traders and investors
Engineering Practices: Developing an Autonomous Stock
could make use of such methods to experience increased Trading System. arXiv preprint arXiv:2303.13216.
profits by investing on the right day at the right place. 14. El Hajj, M., & Hammoud, J. (2023). Unveiling the
Successful implementation as part of this project was Influence of Artificial Intelligence and Machine Learning
carried out wherein five algorithms i.e., Long Short-Term on Financial Markets: A Comprehensive Analysis of AI
Memory and deep Neural Network algorithms were Applications in Trading, Risk Management, and Financial
deployed to create precise predictive models for Operations. Journal of Risk and Financial
application in stock price prediction of twelve prevalent Management, 16(10), 434.
Indian Companies namely, Adani Ports, Asian Paints, 15. Mintarya, L. N., Halim, J. N., Angie, C., Achmad, S., &
Kurniawan, A. (2023). Machine learning approaches in
Axis Bank, Housing Development Finance Corporation
stock market prediction: a systematic literature
Limited (HDFC) Bank, Industrial Credit and Investment review. Procedia Computer Science, 216, 96-102.
Corporation of India (ICICI) Bank, Kotak Bank, 16. Hegazy, O., Soliman, O. S., & Salam, M. A. (2014). A
Hindustan Unilever Limited, Maruti, National Thermal machine learning model for stock market
Power plant Corporation (NTPC), Tata Steel, Tata prediction. arXiv preprint arXiv:1402.7351.
Consultancy Services (TCS) and Titan, after which an 17. Nayak, A., Pai, M. M., & Pai, R. M. (2016). Prediction
elaborate comparative analysis of the performances of the models for Indian stock market. Procedia Computer
algorithms during stock price prediction has been carried Science, 89, 441-449.
out. The stock prices collected were from 2015 to 2021, 18. Ding, G., & Qin, L. (2020). Study on the prediction of
stock price based on the associated network model of
and after this exhaustive research, it can be concluded
LSTM. International Journal of Machine Learning and
that DL algorithms have a substantial edge over simple Cybernetics, 11, 1307-1317.
ML algorithms when it comes to the prediction of time 19. Birant, D., & Işık, Z. (2019, October). Stock market
series data. out of the five chosen algorithms, the Long forecasting using machine learning models. In 2019

Innovative Research Publication 24

 International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

Innovations in Intelligent Systems and Applications

Conference (ASYU) (pp. 1-6). IEEE.
20. Moukalled, M. I. (2019). Automated stock price
prediction using machine learning (Doctoral dissertation).
21. Vanukuru, Kranthi. (2018). Stock Market Prediction
Using Machine Learning. 10.13140/RG.2.2.12300.77448.
22. F.R. Sayyed, R.V. Argiddi, S.S. Apte. (2013).
Collaborative Filtering Recommender System for
Financial Market, International Journal of Engineering
and Advanced Technology (IJEAT).
23. R, Kirubahari & Joe Amali, Miruna. (2021). A Hybrid
Deep Collaborative Filtering Approach for Recommender
Systems. 10.21203/rs.3.rs-651522/v1.
24. Prieto-Torres D.R., Galpin I. (2020). A Virtual Wallet
Product Recommender System Based on Collaborative
Filtering. In: Florez H., Misra S. (eds) Applied
Informatics. ICAI 2020. Communications in Computer
and Information Science, vol. 1277. Springer, Cham.
25. D. Rukiya, T. Africa, S. Nayak, S. Nooreain. (2019).
Recommendation in E-Commerce using Collaborative
Filtering International Research Journal of Engineering
and Technology, 6 (5)

Innovative Research Publication 25