ISSN 2593-7568

RESEARCH ARTICLE
Optimizing Stock Market Forecasts: The Role of AI and
Hybrid Models in Predictive Analytics
Shivani Modi1*, Ved Prakash Upadhyay2
1University of Illinois at Urbana Champaign, Illinois, USA
2Columbia University, NY, USA

Abstract
Forecasting stock market movements is a challenging and significant task for both researchers
and investors. Stock market movements are affected by local and global economic factors, as
well as political developments. This field of research requires substantial knowledge of
finance, statistics, and Artificial Intelligence to achieve reliable results. To understand stock
market movements, we must interpret a significant amount of information from non-linear,
volatile, and non-parametric raw data. To reduce the complexity of stock market forecasting,
we need to extract key features from this raw data. To simplify the task of stock market
forecasting for researchers and traders, we conducted a study on the Indian stock market and
present a comprehensive summary report. This report includes an analysis of 50 research
articles related to the Indian stock market, along with some highly cited articles pertaining to
other international markets.

Key Words: Indian stock market forecasting; Stock market prediction; Neural network; Support vector
machine; Artificial intelligence; ARIMA; Random forest

1. Introduction
The volatile behavior of stock markets has been widely debated for decades. Researchers from
both economics and engineering fields have examined the market using various financial and
soft computing models to predict future trends. They have also developed models to forecast
stock market volatility, focusing on cultivating different approaches to successfully predict
future stock prices and market indices. The main goal of these researchers is to build
predictive models using minimal data while achieving high accuracy. Forecasting is
inherently challenging due to the numerous complexities involved. To predict stock markets
accurately, researchers must select appropriate input variables and modeling techniques, as
well as implement accurate performance measures for their models.

*Corresponding Author: Shivani Modi, University of Illinois at Urbana Champaign, Illinois, USA, E-
mail: sm5060@columbia.edu
Received Date: June 04, 2024, Accepted Date: June 24, 2024, Published Date: June 28, 2024
Citation: Modi S, Upadhyay VP. Optimizing Stock Market Forecasts: The Role of AI and Hybrid Models in
Predictive Analytics. Int J Auto AI Mach Learn. 2024;4(1):61-72.
This open-access article is distributed under the terms of the Creative Commons Attribution Non-
Commercial License (CC BY-NC) (http://creativecommons.org/licenses/by-nc/4.0/), which permits
reuse, distribution and reproduction of the article, provided that the original work is properly cited,
and the reuse is restricted to non-commercial purposes.

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 61

 ISSN 2593-7568

In recent years, researchers have become increasingly interested in using AI and business
models to predict stock market volatility [1]. A stock market index is a statistical indicator
used to measure changes in the market value of stocks, providing information about the
overall movements of different stocks in the market. While stock market forecasting has
become a highly researched field in developed countries like the United States, it remains a
relatively new area of study in developing countries such as India. Indian researchers are still
becoming aware of the field's potential.

The purpose of our research is to review and classify the techniques described in the literature.
We have reviewed 50 highly cited research papers covering both Indian markets and other
reputable international markets. In this paper, we summarize the approaches used in various
stock market forecasting systems. We have produced a comprehensive report on data samples
and sizes, input variables, preprocessing techniques, and classifiers used in stock market
forecasting models. We also compare both simple and hybrid models.

Section II presents various input variables used by researchers, while Section III elaborates on
the complete preprocessing approach applied to raw data. In Section IV, feature indicators are
classified according to financial and macroeconomic fundamentals. Sections V and VI provide
a comparative study of classification models and performance measures, respectively, used
in various forecasting systems. Finally, we conclude our study in Section VIII.

2. Input Variables
Various input variables, such as opening, low, high, and closing values of stocks or stock
market indices, are utilized to predict short-term market movements. Our survey reveals that
the number and type of input variables fluctuate based on the researchers' objectives and the
availability of data. Several Indian researchers [2-9] and international researchers [10-12] have
employed the closing value as the sole input variable for market movement forecasting.

Certain researchers focus on long-term predictions, incorporating macroeconomic variables

such as exports and imports, money supply, interest rates, inflation rates, foreign exchange
rates, unemployment figures, and detailed company financial profiles [13,14]. These profiles
include metrics such as dividend yields, earnings yield, cash flow yield, book-to-market ratio,
price-earnings ratio, lagged returns, and firm size. In specific instances, researchers have
integrated economic and financial factors as input variables in their models [13,15-20].

3. Data Preprocessing
Data preprocessing is a fundamental step in the data mining process, particularly when
forecasting time series such as stock markets. Preprocessing raw data is essential because the
available raw data is often inadequate for modeling purposes due to its highly inconsistent
nature. Raw data from various sources typically contains noise, including redundancy and
missing values, which can significantly affect the quality of the data. The quality of data used
for modeling directly impacts the accuracy of the prediction model. Therefore, data
preprocessing is employed to enhance accuracy and reduce the complexity of the model.

During data preprocessing, data is normalized so that each input component is linearly scaled
within a specified range, such as (-1.0, 1.0) or (0, 1). This normalization ensures that the data
is standardized and comparable. Researchers who scaled their samples within the range of (0,

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 62

 ISSN 2593-7568

1) are referenced in articles [21-24]. In contrast, articles [4,7,12,25,26] discuss researchers who
scaled their samples within the range of (-1, 1). Additionally, some researchers have used
different scaling ranges, such as (-0.9, 0.9) and (-0.5, 0.5), as shown in articles [27,28]
respectively. These variations in scaling methods reflect the diverse approaches researchers
take to optimize their models.

To reduce the dimensionality of data, principal component analysis (PCA) is commonly

employed, as cited by authors of articles [18,29-32]. High dimensionality, caused by
redundancy in source data, is typically summarized from numerous independent variables to
a smaller set of derived variables known as principal components. The principal components
capture the most significant variance in the data. The number of principal components is
always less than or equal to the number of original variables. Principal components are
derived from the variance matrix, with the first principal component being the linear
combination of matrix elements that exhibit the largest possible variance, followed by the
second component, which captures the second-largest variance, and so on. This reduction in
dimensionality helps in simplifying the model and improving its performance.

Time series data is inherently non-stationary and high-dimensional. A time series is

considered stationary if there are no systematic changes in mean (no trend), no changes in
variance, and any periodic variations have been removed. Stationarity is a critical property
for many time series forecasting models, as it simplifies the modeling process and improves
accuracy. To test the stationarity of data, unit root tests such as the Augmented Dickey-Fuller
test [17,33] and the Phillips-Perron test [6,34] are employed. These tests help in determining
whether the time series data needs to be transformed to achieve stationarity, ensuring that the
forecasting model produces reliable and accurate predictions.

4. Feature Extraction
To enhance the accuracy of predictive models, it is essential to extract meaningful features
from the dataset. When the source data is large and potentially redundant, it becomes
necessary to condense the data into a set of significant information, known as feature
indicators. It is expected that these feature indicators, extracted from the raw stock market
time series data, encompass all pertinent and relevant information.

In the literature, various feature indicators are employed, generally categorized into two
types: fundamental and technical, based on stock market analysis. Fundamental analysis
relies on macroeconomic data, including exports and imports, money supply, interest rates,
inflation rates, foreign exchange rates, unemployment figures, and specific company financial
profiles (e.g., dividend yields, earnings yield, cash flow yield, book-to-market ratio, price-
earnings ratio, lagged returns, and company size).

On the other hand, technical analysis typically disregards the efficient market hypothesis,
operating on the rationale that history will repeat itself and that the correlation between price
and volume can reveal market behavior. Predictions are made by exploiting insights hidden
in past trading activities and by analyzing patterns and trends in price and volume charts.

The selection of feature indicators is influenced by the specific time series of the stock market,
known as the forecasting horizon. Feature indicators can be chosen based on the critical factors
affecting the forecasting horizon. If the forecasting time horizon spans one year or more,
fundamental analysis is preferred. Conversely, if the horizon is shorter than one-year,

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 63

 ISSN 2593-7568

technical analysis is more suitable. Since modern forecasting systems can automatically select
relevant features, the exact number of indicators is less critical; only those indicators
containing relevant information will be chosen by the system.

In the literature, researchers commonly classify feature indicators into three types: volume-
based, price-based, and overlay indicators. Volume-based indicators analyze the total trading
volume in stock markets, while price-based indicators examine the stock's price value. In
technical charts, feature indicators often appear as squiggly lines above, below, or on top of
the price information. These are known as overlay indicators, which use the same scale as
prices and are typically plotted on top of the price bars. Table 1 categorizes all feature
indicators used in the literature, providing a comprehensive overview of their application.

5. Classification and Modeling

The choice of a suitable classifier is crucial for enhancing the performance of stock market
forecasting systems, both for intraday and long-term predictions. Researchers have
experimented with various types of forecasting classifiers, utilizing models like ARIMA (Auto
Regressive Moving Average) [2, 6, 25] EGARCH (Exponential Generalized Auto Regressive
Conditional Heteroskedasticity), TARCH (Threshold ARCH) [3,33] Hidden Markov Models
[30,35] and ARFIMA-FIGARCH (Auto Regressive Fractionally Integrated Moving Average-
Fractionally Integrated Generalized Autoregressive Conditional Heteroskedasticity) [23,36].
These models employ statistical techniques for forecasting.

Recently, there has been a significant shift towards machine learning algorithms for stock
market predictions due to their promising results. However, it is essential for researchers to
consider several influential parameters during the modeling process. Surveys indicate that
the most prominent factors affecting share prices are their immediate opening and closing
values [37]. Additionally, the forecasting horizon and investment strategies significantly
influence trading simulations. The forecasting horizon refers to the period during which the
indices are realized [13]. From our study, we have observed that neural networks' ability to
discover nonlinear relationships in input data makes them preferable for modeling nonlinear
dynamic systems like the stock market [38,39,32].

The Artificial Neural Network (ANN) classifier is particularly adept at identifying outliers
and erroneous data [13]. By implementing the Empirical Risk Minimization (ERM) principle,
ANNs often outperform traditional statistical models [11,5]. It is noted that the risk-adjusted
performance of NN-based trading models generally surpasses the Buy and Hold strategy [28].
ANN's capacity to learn nonlinear patterns accurately has led to its widespread acceptance in
stock market prediction [24,40]. Various versions of neural network classifiers are used in
stock market forecasting, as cited in [4,17,21,28,41-43].

Despite their advantages, ANNs are prone to overfitting, local minima traps, and challenges
in determining the hidden layer size and learning rate [11]. These issues can limit their
practical application and reliability in fluctuating market conditions. In comparison, the
Support Vector Machine (SVM) model [44,45] offers good generalization performance,
absence of local minima, and sparse representation of solutions. The SVM classifier is based
on the Structural Risk Minimization (SRM) principle, which differs from the ERM principle
that only minimizes training error. Due to SRM, SVMs often achieve higher generalization
performance than traditional ANNs [10,26].

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 64

 ISSN 2593-7568

Table 1: Feature Indicators.

Indicators Explanation Type

On-Balance OBV is a momentum indicator that uses volume flow to predict changes in stock
Volume (OBV) price
Money Flow Compares the traded value of the up-days to the traded value of down-days
Index (MFI) and puts it in a percentage value. Volume
Volume Price It consists cumulative volume line that adds or subtracts a multiple of the based
Trend percentage change in share price trend and current volume, depending upon indicators
Indicator their upward or downward movements.
Chaikin CMF Oscillator is derived from MACD. Used in technical analysis as
Money Flow measurement of buying and selling pressure with a purpose of generating
trading signals.
Relative A technical momentum indicator that compares the magnitude of recent gains to
Strength Index recent losses in an attempt to determine overbought and oversold conditions of
(RSI) an asset.
Moving The difference between a fast and slow exponential moving average (EMA) of
Average closing prices. (Fast means a short-period average, and slow means a long period
Convergence one)
Divergence
(MACD) Price
Rate Of The percentage difference between the current price and the price n-time periods based
Change (ROC) ago. indicators
Stochastic It compares a security's closing price to its price range over a given time period.
Oscillator (SO) The oscillator's sensitivity to market movements can be reduced by adjusting the
time period or by taking a moving average of the result
William’s %R This is a momentum indicator measuring overbought and oversold levels,
similar to a stochastic oscillator
Momentum To identify trend lines used an oscillator which is known as momentum.
Chaikin Combines price and volume to show how money may be flowing into or out of
Oscillator a stock. Based on Accumulation/Distribution Line
Moving To emphasize the direction of a trend and smooth out price and volume
Average fluctuation that can confuse interpretation. Overlay
Bollinger A chart overlay that shows the upper and lower limits of price movements based based
Bands on the Standard Deviation of prices. indicators

Training an SVM is equivalent to solving a linearly constrained quadratic programming

problem, ensuring a unique and globally optimal solution. Unlike other networks that risk
getting stuck in local minima, SVMs depend only on a subset of training data points, known
as support vectors, which simplifies computation. Literature shows that SVMs outperform
random forests, neural networks, and other traditional models due to their implementation

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 65

 ISSN 2593-7568

of the SRM principle [10,26,46,47]. Various stock market forecasting models have utilized the
SVM classifier, as seen in [12,25,27,31,46,48-53].

Similar to SVM, the regularized Radial Basis Function (RBF) neural network minimizes the
regularized risk function, leading to better generalization performance than the Back
Propagation (BP) neural network [10]. The RBF network's robustness to overfitting makes it a
reliable alternative in fluctuating markets. However, a disadvantage of SVM is that its training
time scales quadratically or cubically with the number of training samples, making it
computationally intensive for large datasets [10]. This limitation necessitates efficient
algorithmic strategies and high computational resources for practical implementation.

Experiments with mixed classifiers have shown better results compared to single classifier
models. For instance, a decision tree-KPCA-ANFIS hybrid system outperforms simple neural
networks and naïve Bayesian models [54,55]. The decision tree rough set-based prediction
system also outperforms standalone rough set and ANN-based systems without feature
selection [29]. A hybrid ARIMA-GARCH model with a Moving Average (MA) filter-based
decomposition pre-processing step outperforms ARIMA, GARCH, trend-ARIMA, and
wavelet-ARIMA models [35]. A hybrid neuro-fuzzy adaptive control system has
demonstrated superior performance compared to 13 other soft computing approaches [14].
Furthermore, models learning through trend deterministic data showed significant
performance improvements, achieving accuracies of 86.69% for ANN, 89.33% for SVM,
89.98% for random forest, and 90.19% for naive-Bayes [25].

The Cuckoo Search (CS) algorithm, based on Swarm Intelligence optimization, effectively
tunes SVM parameters, resulting in higher accuracy rates than regular SVM methods [44].
Experimental results indicate that the CS-SVM method provides higher accuracy with lower
Mean Squared Error (MSE) and Mean Absolute Percentage Error (MAPE) compared to ANN
and SVM. Additionally, a hybrid ARIMA-neural network model [5] outperforms linear
ARIMA and nonlinear ANN models. The Genetic Algorithm (GA) optimized decision tree-
SVM hybrid system [48] outperforms both ANN and naïve Bayes prediction systems, as well
as standalone SVM models [50]. These hybrid models leverage the strengths of individual
algorithms, leading to more robust and accurate forecasting results. In summary, while
various models and hybrid systems have shown promise in stock market forecasting, the
choice of classifier and the consideration of influential parameters are vital for achieving
optimal performance. Continuous advancements in machine learning and hybrid modeling
approaches hold the potential to further enhance the accuracy and reliability of stock market
predictions, aiding investors in making informed decisions.

6. Performance Measures
Performance measures, also known as quality measures, play an indispensable role in the field
of machine learning and data sciences. They are crucial for evaluating the strength of
classification models and serve as criteria for designing heuristics to develop these models.
Performance measures can be broadly classified into two types: statistical and non-statistical
measures.

Statistical measures are further divided into parametric and non-parametric tests. Parametric
tests make certain assumptions about the parameters (properties) of the data distribution.
These assumptions help in defining the nature and behavior of the data, facilitating more

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 66

 ISSN 2593-7568

precise analysis under specified conditions. Examples of parametric tests include t-tests and
ANOVA, which assume normal distribution and equal variances.

On the other hand, non-parametric tests make no assumptions about the data distribution.
They are more flexible and can be used with data that do not fit the assumptions of parametric
tests. Examples of non-parametric tests include the Mann-Whitney U test and the Kruskal-
Walli’s test, which are useful for ordinal data or non-normally distributed data.

Non-statistical performance measures include metrics like the hit ratio, which is widely used
in time series problems. The hit ratio evaluates how often the predicted values correctly match
the actual values, providing insight into the model's accuracy over a period. These
performance measures, whether statistical or non-statistical, are essential for the
development, validation, and refinement of machine learning models, ensuring their
effectiveness and reliability in various applications.

Table 2: Performance Measures [7].

MSE Mean squared error Mean (et^2 )

RMSE Root mean squared error √ (MSE)
MAE Mean absolute error mean(|et|)
Md. AE Media absolute error median(|et|)
MAPE Mean absolute percentage error mean(|pt |)
Md. APE Median absolute percentage error median(|pt|)
SMAPE Symmetric mean absolute percentage error mean(2|Yt - Ft |/( Yt +Ft))

SMd. APE Symmetric median absolute percentage error median(2|Yt -Ft |/( Yt +Ft))

If the problem is a regression problem, performance measures such as Root Mean Squared
Error (RMSE), Mean Absolute Error (MAE), and Steady State Error (SSE) are typically used.
These metrics help quantify the difference between the predicted and actual values, providing
insights into the accuracy and reliability of the regression model.

In classification problems, performance is often evaluated using a confusion matrix or error

matrix. These matrices provide a comprehensive overview of the model's performance by
displaying the counts of true positives, true negatives, false positives, and false negatives.
From these counts, various metrics such as accuracy, precision, recall, and F1 score can be
derived, offering a detailed assessment of the classifier's effectiveness. Table 2 illustrates the
performance measures commonly used in the literature for both regression and classification
problems, highlighting their significance in model evaluation.

7. Conclusion
Our comprehensive survey of intelligent system techniques for Indian stock market
forecasting has revealed significant insights into the efficacy of various models. The research
underscores that artificial intelligence (AI) techniques, particularly machine learning models,
generally outperform traditional statistical methods in predicting stock market trends.
Among the soft computing techniques examined, the Support Vector Machine (SVM) stands
out as the most preferred classification model due to its superior accuracy and generalization
capabilities compared to other models.

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 67

 ISSN 2593-7568

The literature also provides robust evidence supporting the effectiveness of hybrid models
over single classifier models. These hybrid approaches leverage the strengths of multiple
algorithms, leading to enhanced performance and reliability in stock market predictions. For
instance, hybrid models such as ARIMA-GARCH, decision tree-KPCA-ANFIS, and neuro-
fuzzy adaptive control systems have demonstrated superior accuracy and robustness
compared to their standalone counterparts. These findings indicate that combining different
modeling techniques can significantly improve forecasting outcomes.

Additionally, our analysis highlights the importance of data preprocessing and feature
extraction in developing reliable forecasting models. Techniques such as normalization,
principal component analysis (PCA), and the selection of appropriate feature indicators are
critical in enhancing accuracy and reducing the complexity of predictive models. The choice
of input variables, whether fundamental or technical, also plays a pivotal role in the success
of these models.

Despite the promising results of AI and hybrid models, challenges such as overfitting, local
minima traps, and computational intensity remain. Advanced techniques and high
computational resources are essential to address these issues effectively. Moreover,
continuous advancements in machine learning algorithms and hybrid modeling approaches
hold the potential to further enhance the accuracy and reliability of stock market predictions,
providing valuable tools for investors and researchers alike.

In conclusion, our study reaffirms the significant potential of AI and hybrid techniques in
stock market forecasting. By leveraging these advanced methodologies, researchers and
investors can achieve more accurate and reliable predictions, thereby making more informed
decisions in the dynamic and complex world of stock markets. Future research should
continue to explore and refine these techniques, focusing on overcoming current limitations
and harnessing new advancements in the field.

References
1. Devadoss AV, Ligori TA. Stock prediction using artificial neural networks. Int J Data
Min Tech App. 2013;2:283-91.

2. Abraham A, Philip NS, Saratchandran P. Modeling chaotic behavior of stock indices

using intelligent paradigms. ArXiv Prepr. 2004:cs/0405018.

3. Chen AS, Leung MT, Daouk H. Application of neural networks to an emerging

financial market: forecasting and trading the Taiwan stock index. Comput Oper Res.
2003;30:901-23.

4. Khan AU, Bandopadhyaya TK, Sharma S. Genetic algorithm-based backpropagation

neural network performs better than backpropagation neural network in stock rates
prediction. Int J Comput Sci Netw Secur. 2008;8:162-6.

5. Dutta A, Bandopadhyay G, Sengupta S. Prediction of stock performance in the Indian

stock market using logistic regression. Int J Bus Inf. 2012;7:105.

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 68

 ISSN 2593-7568

6. Hiremath GS, Kamaiah B. Some further evidence on behaviour of stock returns in

India. Int J Econ Finance. 2010;2:1-11.

7. Nair BB, Minuvarthini M, Sujithra B, et al. Stock market prediction using a hybrid
neuro-fuzzy system. International conference on advances in recent technologies in
communication and computing, Kottayam, India. 2010.

8. Nair BB, Mohandas VP, Sakthivel NR. A decision tree-rough set hybrid system for
stock market trend prediction. Int J Comput Appl. 2010;6:1-6.

9. Babu CN, Reddy BE. Selected Indian stock predictions using a hybrid ARIMA-
GARCH model. International conference on advances in electronics computers and
communications, Bangalore, India. 2014.

10. Cao LJ, Tay FE. Support vector machine with adaptive parameters in financial time
series forecasting. IEEE Trans Neural Netw. 2003;14:1506-18.

11. Yeh CY, Huang CW, Lee SJ. A multiple-kernel support vector regression approach for
stock market price forecasting. Expert Syst Appl. 2011;38:2177-86.

12. Grosan C, Abraham A, Ramos V, et al. Stock market prediction using multi expression
programming. Portuguese conference on artificial intelligence, Covilha, Portugal.
2005.

13. Upadhyay VP, Panwar S, Merugu R. Protein sequence structure prediction using
artificial intelligent techniques. Proceedings of the international conference on
advances in information communication technology & computing, Bikaner, India.
2016.

14. Banerjee D. Forecasting of Indian stock market using time-series ARIMA model. 2nd
international conference on business and information management, Durgapur, India.
2014.

15. Atsalakis GS, Valavanis KP. Forecasting stock market short-term trends using a neuro-
fuzzy based methodology. Expert Syst Appl. 2009;36:10696-707.

16. Zhang G, Patuwo BE, Hu MY. Forecasting with artificial neural networks: the state of
the art. Int J Forecast. 1998;14:35-62.

17. Kaur H. Time varying volatility in the Indian stock market. Vikalpa. 2004;29:25-42.

18. Goudarzi H, Ramanarayanan CS. Modeling asymmetric volatility in the Indian stock
market. Int J Bus Manag. 2011;6:221.

19. Patel J, Shah S, Thakkar P, et al. Predicting stock and stock price index movement using
trend deterministic data preparation and machine learning techniques. Expert Syst
Appl. 2015;42:259-68.

20. Badge J. Forecasting of Indian stock market by effective macro-economic factors and
stochastic model. J Stat Econom Methods. 2012;1:39-51.

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 69

 ISSN 2593-7568

21. Devi KN, Bhaskaran VM, Kumar GP. Cuckoo optimized SVM for stock market
prediction. International conference on innovations in information, embedded and
communication systems, Coimbatore, India. 2015.

22. Sujatha KV, Sundaram SM. A combined PCA-MLP model for predicting stock index.
Proceedings of the 1st Amrita ACM-W celebration on women in computing in India,
Coimbatore, India. 2010.

23. Vaisla KS, Bhatt AK. An analysis of the performance of artificial neural network
technique for stock market forecasting. Int J Comput Sci Eng. 2010;2:2104-9.

24. Kim KJ. Financial time series forecasting using support vector machines.
Neurocomputing. 2003;55:307-19.

25. Dey L, Mahajan A, Haque SM. Document clustering for event identification and trend
analysis in market news. Seventh International Conference on Advances in Pattern
Recognition, Kolkata, India. 2009.

26. Kumar M, Thenmozhi M. Support vector machines approach to predict the S&P CNX
NIFTY index returns. J Emerg Mark Finance. 2007:1-19.

27. Rajakumar MP, Shanthi V. Computationally intellectual structure for forecasting share
price. Int J Soft Comput. 2013;8:218-22.

28. Thenmozhi M. Forecasting stock index returns using neural networks. Delhi Business
Review. 2006;7:59-69.

29. Subha MV, Nambi ST. Classification of stock index movement using k-nearest
neighbours (k-NN) algorithm. WSEAS Trans Inf Sci Appl. 2012;9:261-70.

30. Kumar M. A time-varying parameter vector autoregression model for forecasting

emerging market exchange rates. Int J Econ Sci Appl Res. 2010;3:21-39.

31. Hassan MR, Nath B, Kirley M. A fusion model of HMM, ANN and GA for stock
market forecasting. Expert Syst Appl. 2007;33:171-80.

32. Nair BB, Mohandas VP, Sakthivel NR. A genetic algorithm optimized decision tree-
SVM based stock market trend prediction system. Int J Comput Sci Eng. 2010;2:2981-
8.

33. Dutta G, Jha P, Laha AK, et al. Artificial neural network models for forecasting stock
price index in the Bombay stock exchange. J Emerg Mark Finance. 2006;5:283-95.

34. Merh N, Saxena VP, Pardasani KR. A comparison between hybrid approaches of ANN
and ARIMA for Indian stock trend forecasting. Bus Intell J. 2010;3:23-43.

35. Upadhyay VP, Sonawat M, Singh S, et al. Nano robots in medicine: a review. Int J Eng
Technol Manag Res. 2017;4:27-37.

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 70

 ISSN 2593-7568

36. Sivakumar PB, Mohandas VP. Modeling and predicting stock returns using the
ARFIMA-FIGARCH. World congress on nature & biologically inspired computing,
Coimbatore, India. 2009.

37. Chang PC, Liu CH. A TSK type fuzzy rule based system for stock price prediction.
Expert Syst Appl. 2008;34:135-44.

38. Troiano L, Kriplani P. Predicting trend in the next-day market by Hierarchical Hidden
Markov Model. International conference on computer information systems and
industrial management applications, Karakow, Poland. 2010.

39. Choudhry R, Garg K. A hybrid machine learning system for stock market forecasting.
Int J Comput Inform Eng. 2008;2:689-92.

40. Nayak RK, Mishra D, Rath AK. A Naïve SVM-KNN based stock market trend reversal
analysis for Indian benchmark indices. Appl Soft Comput. 2015;35:670-80.

41. Merugu R, Upadhyay VP, Manda S. Bioinformatics analysis and modelling of

mycotoxin patulin induced proteins. Int J Bioinformatics Biol Sci. 2016;4:5-10.

42. Dhar S, Mukherjee T, Ghoshal AK. Performance evaluation of neural network

approach in financial prediction: evidence from Indian market. International
Conference on Communication and Computational Intelligence, Erode, India. 2010.

43. Das SP, Padhy S. Support vector machines for prediction of futures prices in Indian
stock market. Int J Comput Appl. 2012;41:22-6.

44. Mitra SK. Optimal combination of trading rules using neural networks. Int Bus Res.
2009;2:86-9.

45. Chaudhuri TD, Ghosh I. Using clustering method to understand Indian stock market
volatility. ArXiv Prepr. 2016:1604.05015.

46. Huang W, Nakamori Y, Wang SY. Forecasting stock market movement direction with
support vector machine. Comput Oper Res. 2005;32:2513-22.

47. Kara Y, Boyacioglu MA, Baykan OK. Predicting direction of stock price index
movement using artificial neural networks and support vector machines: the sample
of the Istanbul stock exchange. Expert Syst Appl. 2011;38:5311-9.

48. Chen Y, Dong X, Zhao Y. Stock index modeling using EDA based local linear wavelet
neural network. International conference on neural networks and brain, Beijing,
China. 2005.

49. Chen Y, Abraham A, Yang J, et al. Hybrid methods for stock index modeling. Fuzzy
systems and knowledge discovery: second international conference, Changsha, China,
2005.

50. Kumar M, Thenmozhi M. Stock index return forecasting and trading strategy using
hybrid ARIMA-neural network model. Int J Finance Manag. 2012;2:1-14.

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 71

 ISSN 2593-7568

51. Handa R, Hota HS, Tandan SR. Stock market prediction with various technical
indicators using neural network techniques. Int J Res Appl Sci Eng Technol. 2015;3:604-
8.

52. Atsalakis GS, Valavanis KP. Surveying stock market forecasting techniques–Part II:
soft computing methods. Expert Syst Appl. 2009;36:5932-41.

53. Kumar M, Thenmozhi M. Forecasting stock index movement: a comparison of support

vector machines and random forest. Indian institute of capital markets 9th capital
markets conference, Hong Kong, China. 2006.

54. Upadhyay VP, Panwar S, Merugu R, et al. Forecasting stock market movements using
various kernel functions in support vector machine. Proceedings of the international
conference on advances in information communication technology & computing,
Bikaner, India. 2016.

55. Panwar S, Upadhyay VP, Bishnoi SK. A survey of intelligent system techniques for
Indian stock market forecasting. 39th MIPRO conference, Opatija, Croatia. 2016.

Int J Auto AI Mach Learn, Vol 4, Issue 1, June 2024 72