International Journal of Electrical and Computer Engineering (IJECE)

Vol. 15, No. 2, April 2025, pp. 2192∼2201

ISSN: 2088-8708, DOI: 10.11591/ijece.v15i2.pp2192-2201 ❒ 2192

A novel technique for selecting financial parameters and

technical indicators to predict stock prices
Sneha S. Bagalkot1,2 , Dinesha H. A.3 , Nagaraj Naik4
1 Nagarjuna College of Engineering and Technology, Bengaluru and Visvesvaraya Technological University, Belagavi, India
2 Department of Computer and Science Engineering, B.M.S. College of Engineering, Bangalore, India
3 Department of Computer and Science Engineering, Shridevi Institute of Engineering and Technology, Tumkuru, India
4 Department of Computer Science and Engineering, Manipal Institute of Technology, Manipal Academy of Higher Education Manipal,

Manipal, India

Article Info ABSTRACT

Article history: Stock price predictions are crucial in financial markets due to their inherent
volatility. Investors aim to forecast stock prices to maximize returns, but accu-
Received May 30, 2024
rate predictions are challenging due to frequent price fluctuations. Most litera-
Revised Nov 13, 2024 ture focuses on technical indicators, which rely on historical data. This study
Accepted Nov 20, 2024 integrates both financial parameters and technical indicators to predict stock
prices. It involves three main steps: identifying essential financial parameters us-
Keywords: ing recursive feature elimination (RFE), selecting quality stocks with a decision
tree (DT), and forecasting stock prices using artificial neural networks (ANN),
Artificial neural network deep neural networks (DNN), and extreme gradient boosting (XGBoost). The
Deep neural network models’ performance is evaluated with root mean square error (RMSE) and
Financial parameters mean absolute error (MAE) scores. ANN and DNN models showed superior
Recursive feature elimination performance compared to the XGBoost model. The experiments utilized Indian
XGBoost stock data.

This is an open access article under the CC BY-SA license.

Corresponding Author:
Ngaaraj Naik
Department of Computer Science and Engineering, Manipal Institute of Technology, Manipal Academy of
Higher Education
Manipal-576 104, India
Email:[email protected]

1. INTRODUCTION
The stock market attracts numerous investors and traders due to its potential for rapid and substantial
gains [1], [2]. Both institutional and individual investors engage in stock trading [3], [4]. Institutional investors,
being trained professionals, manage significant investment portfolios [5], [6] and typically focus on long-term
investments. In contrast, retail investors often lack the trading expertise necessary for optimal buy or sell
decisions [7]. Consequently, a robust stock forecasting model is essential for assisting retail investors.
To identify stock price trends, investors utilize two primary methods of analysis. The first is technical
analysis, which assesses stock trends using historical data such as opening, closing, high, and low prices, along
with trading volume. While technical analysis is applicable for both short-term and long-term investments,
its limitation lies in its exclusive reliance on historical data, often overlooking fundamental factors crucial for
future price predictions [8]. This study, therefore, integrates both financial parameters and historical stock data
to enhance forecasting accuracy.

Journal homepage: http://ijece.iaescore.com

Int J Elec & Comp Eng ISSN: 2088-8708 ❒ 2193

The fundamental approach involves analyzing financial statements [9] and evaluating metrics like
earnings, assets, liabilities, and book value [10], [11]. The challenge arises from the plethora of financial
parameters, making it difficult to determine which ones are most predictive. To address this, the study employs
the recursive feature elimination (RFE) technique to identify the most relevant financial parameters.
Subsequently, the study uses a decision tree method to identify high-quality stocks. Decision trees
are effective for classifying target variables [12], [13], as they partition data based on features, with nodes
representing features and leaf nodes indicating outcomes based on entropy reduction or information gain. After
identifying quality stocks, the study predicts their future prices using data mining methods, particularly machine
learning [14],[15]. Machine learning approaches, including supervised and unsupervised methods [16], are
employed, with artificial neural networks (ANNs) being used for their ability to handle non-linear and volatile
data [17], [18]. Additionally, boosting algorithms, particularly XGBoost, are gaining traction for their gradient
boosting ensemble approach [19], providing scalable solutions for end-to-end tree boosting. Although many
studies focus solely on historical stock data for forecasting [20]–[22], integrating financial parameters with
historical data is crucial for accurate predictions. This study, therefore, combines these elements to improve
the estimation of future stock prices.
The contributions of this paper are summarized as follows:
a. We considered the RFE technique to identify the important financial parameters.
b. To identify the quality of the stock, we considered the decision tree.
c. Financial parameters and historical stock data are considered to forecast the future value of the stock price
using ANN, deep neural networks (DNN), and XGBoost models.
Section 2 describes literature reviews, and section 3 provides methodology. Results are presented in section 4.
Section 5 presents the conclusion of the work and future directions.

2. LITERATURE REVIEWS
Yun et al. [23] introduced the Genetic-XGBoost algorithm with a three-stage feature engineering
method. Their model, incorporating 67 technical indicators, achieved an average prediction accuracy of 93.28%
on the KOSPI dataset, significantly outperforming previous models. The study emphasizes the importance
of selecting an optimal feature set to avoid overfitting and simplify interpretation. Gunduz et al. [24] used
LightGBM and long short-term memory (LSTM) classifiers to predict banking stock movements on the Borsa
Istanbul exchange. The LSTM models utilized 2D tensors, incorporating data from the previous eight hours for
hourly predictions, demonstrating the effectiveness of temporal data integration. Peng et al. [25] considered
124 technical indicators for stock price prediction, employing feature selection techniques to eliminate irrele-
vant indicators. They used a deep neural network model and fine-tuned hyperparameters, such as hidden layers
and dropout rates, to enhance prediction accuracy.
Li et al. [26] proposed a deep learning framework for gross domestic product (GDP) prediction involv-
ing three steps: feature extraction with feature crossing, relationship analysis with Boruta-RF and Q-learning,
and model creation with TCN. Their method improved temporal convolutional network (TCN) prediction ac-
curacy by 10%. Selecting the best subset of features is NP-hard, necessitating heuristic methods [27], [28].
Feature selection’s scalability is crucial for large datasets, yet high-dimensional data can lead to local optima
and costly parameter adjustments.
Handling non-linear and noisy data for feature selection is challenging [29]. Recent studies show that
deep learning (DL) models, particularly LSTM, excel in financial forecasting by leveraging long-term depen-
dencies [30], [31]. Time complexity can be reduced by eliminating variables before model integration. Filter
feature selection scores each feature, while the wrapper technique uses subsets for model training. Combin-
ing these approaches, the genetic algorithm optimizes feature subsets via natural selection. Identifying market
styles is vital for predicting stock prices [32]. The summary of literature reviews are described in Table 1.
Chun et al. [33] developed a stock forecasting model integrating investor emotions by analyzing
microblogging data. The model employs part-of-speech (POS) tagging to extract emotional terms such as
adjectives, nouns, adverbs, and interjections, which are classified into emotions like joy and sadness. While
technical indicators are commonly used for stock price forecasting, they primarily rely on historical data, often
neglecting fundamental company factors. Evaluating stock quality requires considering financial parameters,
which provide insights into future stock value.

A novel technique for selecting financial parameters and technical indicators to ... (Sneha S. Bagalkot)
2194 ❒ ISSN: 2088-8708

Table 1. Literature reviews on stock price prediction

Author Data Method Target output Gap
Yun et al. [23] Technical indicators XGBoost Prediction Stock historical data is considered.
Gunduz et al. [24] Technical indicators LSTM Classification Stock historical data is considered.
Peng et al. [25] Technical indicators Deep neural network Prediction Stock historical data is considered.
Li et al. [26] GDP Deep learning Prediction Stock historical data is considered.
Chen et al. [30] Technical indicators Deep learning Prediction Stock historical data is considered.
Kim et al. [1] Technical indicators Genetic algorithm Prediction Stock historical data is considered.
Kumar et al. [34] Technical indicators ENN Prediction Stock historical data is considered.
Chun et al. [33] Micro-blogging data Deep learning Prediction Stock historical data is considered.

3. PROPOSED METHODOLOGY
The overall work is described in Figure 1. The proposed work is classified into three tasks. The first
is the identification of important financial parameters, the second is identifying quality stock, and the third
is estimating the future value of stock price. The financial parameters are collected from NSE. However, it
consists of many financial parameters such as price to earnings, book value, and returns on equity.. These
financial parameters help identify the intrinsic value of stock prices. In this work, we considered 14 financial
parameters. The aim is to identify important financial parameters. Therefore, we considered the RFE approach
to remove the noisy financial parameters. The next step is to identify the quality stock; therefore decision tree
is considered. In DT, the accuracy metric is used to identify stock quality. Finally, selected quality stocks
extracted historical data and given input to the ANN, DNN, and XGBoost models to forecast the future stock
price.

Financial Financial Accuracy

Identifying
NSE Financial parameters parameters metric is used
quality of stock
parameters selection using selected using to evaluate
using DT
RFE MAE Score quality of stock

ANN
Historical
stock price
collected from DNN
NSE

XGBoost

Figure 1. Financial parameters based stock price prediction

3.1. Recursive feature elimination (RFE)

The RFE method identifies significant stock financial parameters by sequentially eliminating features
based on their MAE scores. This study evaluated 14 financial parameters using random forest regression
to assess feature importance, with earning per share (EPS) as the dependent variable. The list of financial
parameters are described in Table 2. Features with MAE scores exceeding 20% were deemed irrelevant; hence,
other income, depreciation, and dividend payout were excluded due to their high MAE scores. The remaining
parameters were used as inputs for the decision tree to determine stock quality. The step by step are described
in algorithm 1.

3.2. Decision tree (DT)

DT is a common approach for solving regression and classification tasks [35], [36]. The relevant
financial parameters selected by the RFE method are given to DT for classification. The 11 financial parameters

Int J Elec & Comp Eng, Vol. 15, No. 2, April 2025: 2192-2201
Int J Elec & Comp Eng ISSN: 2088-8708 ❒ 2195

are given input to the DT to identify quality stock. The supervised learning method is used in DT. Earning per
share (EPS) is considered the target variable to classify the quality stock. Here more than 20% EPS returns are
considered good quality stock; otherwise, it is classified as no category. The decision tree determines whether
the stock has quality, i.e., yes or no, based on the financial parameters. A DT is used to classify the stock,
either quality stock or not. These yes and no labels are maps in decision tree classifier vectors. The decision
tree simple IF and ELSE conditions are mapped on classifier vectors. A decision tree is generated based on
the input vector data, and the information gain metric is used to split the decision tree. In DT, 70% of data
is considered training, and 30% of data is considered for testing. We have considered the accuracy, precision,
and recall metric to evaluate the decision tree. Using a decision tree, we found that stock with more than 75%
accuracies is considered good quality stock, shown in Table 3. Later these stocks are given input to the ANN,
DNN, and XGBoost for stock price forecasting.

Algorithm 1. Recursive feature wlimination (RFE) algorithm

1: Input:
2: X: Feature matrix (including financial parameters and technical indicators)
3: y: Target variable vector
4: estimator: Machine learning estimator
5: n f eatures: Desired number of features to select
6:
7: Output:
8: Selected feature subset
9: Initialize a list f eatures with indices of all the features in X
10: Initialize current score to a very large number
11: while length of f eatures ¿ n f eatures do
12: Fit the estimator model on X[:, f eatures] and y
13: Calculate the importance of each feature using the estimator model
14: Rank the features based on their importance scores in ascending order
15: Remove the least important feature from f eatures
16: Fit the estimator model on the reduced feature set and calculate the Mean Absolute Error (MAE)
17: if new MAE score is lower than current score then
18: Update current score to the new MAE score
19: else
20: Break the loop
21: end if
22: end while
23: return f eatures

Table 2. Financial parameters MAE score using RFE

Serial No Financial parameter feature MAE Score
1 Sales 18.5
2 Expenses 19.6
3 Operating Profit 17.7
4 Other Income 23.4
5 Depreciation 25.7
6 Interest 16.7
7 Profit before tax 19.8
8 Tax 17.8
9 Net profit 18.7
10 EPS 17.6
11 Price to earning 16.3
12 Price 17.1
13 Dividend Payout 26.2
14 OPM 19.1

A novel technique for selecting financial parameters and technical indicators to ... (Sneha S. Bagalkot)
2196 ❒ ISSN: 2088-8708

Table 3. Decision tree to evaluate quality stock

Stock Accuracy Precision Recall
SunPharma 77.50 0.85 0.77
Cipla 76.30 0.81 0.70
Bharti-Airtel 78.50 0.86 0.77
Bajaj Auto 76.80 0.84 0.74
Federal Bank 77.50 0.82 0.79
RBL Bank 79.60 0.87 0.78
Bank of Baroda 76.50 0.81 0.79

3.3. Artificial neural network (ANN)

ANN was developed based on a biological neuron model that mimics human behavior [37]. ANN has
neurons interconnected with different layers. Using a decision tree, we have identified the quality stock. These
quality stock historical data have been collected. It consists of five variables open price, low price, high price
,volume, and close price. Here the close price is considered the target variable for the ANN model. In ANN,
each neuron performs the computational task [38]. In this work, we considered neural networks, which have
one hidden layer between the input and the output layer. The input data is multiplied with random weight along
with bias. To deal with the nonlinearity in data sigmoid activation function is considered in the ANN model.
The loss function goal is to reduce the error rate in the model. The back-propagation algorithm computes the
gradient of the loss function by randomly assigning weight to neurons in each iteration.
3.4. Extreme gradient boosting (XGBoost)
XGBoost is one of the popular methods to deal with the non-linearity in data [39]. Using a decision
tree, we have identified the quality stock. The next task is forecasting the future value of stock prices. For
that, we considered historical stock data as input to the forecasting model, such as open price, low price, high
price, volume, and close price. Here the close price is considered the target variable for the XGBoost model.
Using the XGBoost method, computed the residual differences between actual and predicted [40]. Based on
the residuals regression tree is constructed in the proposed work. The similarity score of each tree is computed
based on the the (1) and (2).
residualsum, square
Similarity − score = (1)
residual + λ
residualsum
Output − value = (2)
residual + λ

The loss function goal is to reduce the residual error in the tree. We considered the square of residuals and
adjusted the parameter value of λ . The output value will decrease if λ is more than zero. Because we are
maximizing the value produced by the first tree, lambda is reset to zero.
3.5. Deep neural network (DNN)
The deep neural network is one of the popular methods for stock price forecasting. Therefore, this
work considered DNN to forecast the future stock price. Using a decision tree, we have identified the quality
stock. These quality stock historical data have been collected. It consists of five variables open price, low price,
high price, volume, and close price. Here the close price is considered the target variable for the DNN model.
The input layer reads input data and performs the computational task. The output layer produces the output
[41], [42]. The proposed work constructed four hidden layers in DNN architecture. A DNN with four hidden
layers enhances stock price prediction by progressively capturing complex, non-linear relationships within the
data, with each layer refining the patterns to improve accuracy while reducing overfitting. The DNN Equation
is defined in (3).
n
X
Yi = (Bias + wk ∗ A(x)) (3)
k=1

Here wk represents the weights, and A(x) represents the input vector. To deal with the nonlinearity in the

Int J Elec & Comp Eng, Vol. 15, No. 2, April 2025: 2192-2201
Int J Elec & Comp Eng ISSN: 2088-8708 ❒ 2197

data, we considered the rectified linear units activation functions. Backpropagation is used to determine the
appropriate value for weights and biases. We have fine-tuned the DNN parameters, such as the number of
neurons and hidden layers, and adjusted the learning rate to get the best results.

4. RESULTS AND DISCUSSION

In the experiment setup, national stock exchange (NSE) data were used for investigation (https :
//www.nseindia.com/). The proposed work goal is to identify the relevant financial parameters and forecast
the stock price. The proposed work is divided into three tasks. The first is the recursive feature elimination
(RFE) method to identify the relevant financial parameters. Using the RFE method, we identified the impor-
tant financial parameters. Out of 14 features, three features are considered irrelevant features. The irrelevant
features are other income, depression, and dividend payout. The financial parameters with the highest MAE
are considered irrelevant indicators. The second is to identify the quality stock using the decision tree. The de-
cision tree determines whether the stock has quality, i.e., yes or no, based on the accuracy metric. The accuracy
metric is used to assess the stock quality.
Machine learning models are evaluated based on accuracy metrics. The third is stock price forecasting.
The historical stock price is collected for further analysis. It includes stock data such as the open, high, low,
and close prices and the stock price volume. Around 4910 trade sessions were considered for experiments from
May 1, 2007, to October 11, 2022. To forecast the future value of stock price, we considered ANN, DNN, and
XGBoost models. To get the best performance in prediction models, we have varied the number of neurons,
hidden layer, and learning rate.
n
1 X
M AE = ( ) |yi − xi | (4)
n i=1
v
u n
u 1 X
RM SE = t( ) (yi − xi )2 (5)
n i=1

The MAE and RMSE metric is defined in (4) and (5). Figure 2 shows the stock prices predicted using deep
neural network: Figure 2(a) Sun Pharma Stock and Figure 2(b) Cipla Stock. Figure 3 shows the stock prices
predicted using deep neural network: Figure 3(a) Bharti Airtel Stock and Figure 3(b) Federal Bank Stock. Here
red line indicates the predicted price, and the black line indicates the actual price. We found that the predicted
price needs to fit volatile stock prices better. For that, we have fine-tuned the parameters of a model, like the
number of neurons and learning rate. When setting up a DNN for stock prediction, we carefully tuned various
hyperparameters to enhance model performance. We balanced fast convergence and training stability with a
0.01 learning rate. We balanced processing speed and consistent gradient updates using a 32-batch size. We
trained the model over 100 epochs to prevent overfitting and ensure learning. We used the Adam optimizer
to handle sparse gradients and automatically modify the learning rate during training. We reduced overfitting
by randomly deactivating neurons during training with a dropout rate of 0.5. The rectified linear units (ReLU)
activation function added non-linearity to the buried layers. The network had four hidden layers, with neuron
counts decreasing from 128 to 64 to 32. As indicated in Table 4, the DNN outperformed both the ANN and
XGBoost models in most scenarios. For example, in forecasting SunPharma stock, the DNN achieved an MAE
of 3.210941, RMSE of 7.845115, and an R-squared value of 98.60. In contrast, the ANN had an MAE of
3.937627, RMSE of 8.452742, and an R-squared value of 98.20, while the XGBoost model recorded an MAE
of 5.507795, RMSE of 8.986556, and an R-squared value of 97.80. Comparable results were observed for
Cipla, Bharti-Airtel, Federal Bank, RBL Bank, and Bank of Baroda, with the DNN consistently delivering
lower MAE and RMSE values along with higher R-squared values, signifying superior predictive accuracy and
overall model performance. Additionally, these outcomes were benchmarked against existing literature by [43],
where their multilayer perceptron (MLP) and convolutional neural network (CNN) models exhibited higher
MAE and RMSE values, highlighting the effectiveness of our DNN configuration. For instance, their CNN
model’s performance for stock prediction resulted in an MAE of 25.66 and an RMSE of 36.87, showcasing the
superior accuracy of our DNN approach.

A novel technique for selecting financial parameters and technical indicators to ... (Sneha S. Bagalkot)
2198 ❒ ISSN: 2088-8708

(a)

(b)
Figure 2. Stock price predicted using DNN (a) Sun Pharma Stock and (b) Cipla Stock

(a)

(b)
Figure 3. Stock price predicted using DNN (a) Bharti Airtel Stock and (b) Federal Bank Stock

Int J Elec & Comp Eng, Vol. 15, No. 2, April 2025: 2192-2201
Int J Elec & Comp Eng ISSN: 2088-8708 ❒ 2199

Table 4. Result
Stock Predictions Techniques MAE RMSE R-squared
SunPharma ANN 3.937627 8.452742 98.20
SunPharma DNN 3.210941 7.845115 98.60
SunPharma XGBoost 5.507795 8.986556 97.80
Cipla ANN 3.174641 5.191448 98.95
Cipla DNN 3.346246 5.445911 98.85
Cipla XGBoost 5.251966 7.658853 98.50
Bharti-Airtel ANN 6.334113 10.11714 97.30
Bharti-Airtel DNN 7.088787 11.96207 97
Bharti-Airtel XGBoost 5.490016 7.15828 98.30
Bajaj Auto ANN 25.35 28.33 89.60
Bajaj Auto DNN 28 30 88.60
Bajaj Auto XGBoost 33 35 86.60
Federal Bank ANN 18.17 18.98 93.20
Federal Bank DNN 14.50 14.90 94.40
Federal Bank XGBoost 17 17.60 94.10
RBL Bank ANN 21 21.30 91.10
RBL Bank DNN 17 17.60 92.40
RBL Bank XGBoost 18.50 18.60 93.40
Bank of Baroda ANN 22 22.70 91.60
Bank of Baroda DNN 19 19.40 93.10
Bank of Baroda XGBoost 20.10 20.80 92.40
Lu et al. [43] MLP 31.49 39.26 96.99
Lu et al. [43] CNN 25.66 36.87 97.35
Lu et al. [43] BiLSTM 21.95 31.62 98

5. CONCLUSION
The proposed work encompasses three main tasks aimed at enhancing stock price prediction to as-
sist investors in making profitable decisions. First, we identified relevant financial parameters using the RFE
method. Our findings indicated that parameters such as other income, depreciation, and dividend payouts do
not significantly contribute to stock price prediction, as evidenced by MAE scores greater than twenty, reflect-
ing high data variance. Second, we utilized DT to identify quality stocks, providing a basis for robust stock
selection. Lastly, we forecasted stock prices using ANN, DNN, and XGBoost methods. As demonstrated in
experimental work ANN and DNN outperformed XGBoost in most cases due to their ability to handle nonlin-
earity in stock price data. Specifically, the DNN achieved lower MAE and RMSE values and higher R-squared
values across various stocks, indicating superior predictive accuracy. In contrast, XGBoost was less effective
in scenarios with high data volatility. Future research could explore integrating both technical and fundamental
indicators to improve long-term stock price predictions. Additionally, leveraging options and derivatives data
may enhance the accuracy of future stock value forecasts.

REFERENCES
[1] M. Kim, “A data mining framework for financial prediction,” Expert Systems with Applications, vol. 173, p. 114651,
2021, doi: 10.1016/j.eswa.2021.114651.
[2] S. Gao, Q. Meng, and K. C. Chan, “IPO pricing: do institutional and retail investor sentiments differ?,” Economics
Letters, vol. 148, pp. 115–117, 2016, doi: 10.1016/j.econlet.2016.09.029.
[3] A. Kumar and C. M. C. Lee, “Retail investor sentiment and return comovements,” Journal of Finance, vol. 61, no. 5,
pp. 2451–2486, 2006, doi: 10.1111/j.1540-6261.2006.01063.x.
[4] W. Li and S. S. Wang, “Daily institutional trades and stock price volatility in a retail investor dominated emerging
market,” Journal of Financial Markets, vol. 13, no. 4, pp. 448–474, 2010, doi: 10.1016/j.finmar.2010.07.003.
[5] F. Wen, L. Xu, G. Ouyang, and G. Kou, “Retail investor attention and stock price crash risk: Evidence from China,”
International Review of Financial Analysis, vol. 65, p. 101376, 2019, doi: 10.1016/j.irfa.2019.101376.
[6] E. L. Von Thadden, “Long-term contracts, short-term investment and monitoring,” Review of Economic Studies, vol.
62, no. 4, pp. 557–575, 1995, doi: 10.2307/2298077.
[7] T. Vo-Van, H. Che-Ngoc, N. Le-Dai, and T. Nguyen-Trang, “A new strategy for short-term stock investment using
Bayesian approach,” Computational Economics, vol. 59, pp. 887–911, 2022, doi: 10.1007/s10614-021-10115-8.
[8] B. Itri, Y. Mohamed, B. Omar, E. M. Latifa, M. Lahcen, and O. Adil, “Hybrid machine learning for stock price
prediction in the Moroccan banking sector,” International Journal of Electrical and Computer Engineering (IJECE),
vol. 14, no. 3, pp. 3197–3207, 2024, doi: 10.11591/ijece.v14i3.pp3197-3207.

A novel technique for selecting financial parameters and technical indicators to ... (Sneha S. Bagalkot)
2200 ❒ ISSN: 2088-8708

[9] C. F. Lee, “Fundamental analysis, technical analysis, and mutual fund performance,” in Handbook of Financial
Econometrics, Mathematics, Statistics, and Machine Learning, World Scientific, 2021, pp. 3001–3058.
[10] G. D. Abuselidze and A. N. Slobodianyk, “Value assessment of shares of corporate issuers by applying the methods
of fundamental analysis in the stock exchange market,” in Lecture Notes in Networks and Systems, vol. 206, Springer,
2021, pp. 25–39. doi: 10.1007/978-3-030-72110-7 3.
[11] J. Guerard and A. Mark, “Earnings forecasts and revisions, price momentum, and fundamental data: Further explo-
rations of financial anomalies,” Handbook of Financial Econometrics, Mathematics, Statistics, and Machine Learning
(In 4 Volumes), vol. 5, pp. 1151–1209, 2020, doi: 10.1142/9789811202391 0030.
[12] M. S. Saeed et al., “An efficient boosted C5.0 decision-tree-based classification approach for detecting non-technical
losses in power utilities,” Energies, vol. 13, no. 12, p. 3242, 2020, doi: 10.3390/en13123242.
[13] M. C. Wu, S. Y. Lin, and C. H. Lin, “An effective application of decision tree to stock trading,” Expert Systems with
Applications, vol. 31, no. 2, pp. 270–274, 2006, doi: 10.1016/j.eswa.2005.09.026.
[14] A. Kurani, P. Doshi, A. Vakharia, and M. Shah, “A comprehensive comparative study of artificial neural network
(ANN) and support vector machines (SVM) on stock forecasting,” Annals of Data Science, vol. 10, no. 1, pp.
183–208, 2023, doi: 10.1007/s40745-021-00344-x.
[15] A. M. Priyatno, W. F. R. Sudirman, and R. J. Musridho, “Feature selection using non-parametric correlations and
important features on recursive feature elimination for stock price prediction,” International Journal of Electrical
and Computer Engineering (IJECE), vol. 14, no. 2, pp. 1906–1915, 2024, doi: 10.11591/ijece.v14i2.pp1906-1915.
[16] M. Alloghani, D. Al-Jumeily, J. Mustafina, A. Hussain, and A. J. Aljaaf, “A systematic review on supervised and
unsupervised machine learning algorithms for data science,” in Supervised and Unsupervised Learning for Data
Science, Springer, 2020, pp. 3–21. doi: 10.1007/978-3-030-22475-2 1.
[17] G. S. Atsalakis and K. P. Valavanis, “Surveying stock market forecasting techniques - Part II: Soft computing meth-
ods,” Expert Systems with Applications, vol. 36, pp. 5932–5941, 2009, doi: 10.1016/j.eswa.2008.07.006.
[18] L. A. Laboissiere, R. A. S. Fernandes, and G. G. Lage, “Maximum and minimum stock price forecasting of Brazilian
power distribution companies based on artificial neural networks,” Applied Soft Computing Journal, vol. 35, pp.
66–74, 2015, doi: 10.1016/j.asoc.2015.06.005.
[19] J. Pesantez-Narvaez, M. Guillen, and M. Alcañiz, “Predicting motor insurance claims using telematics
data—XGboost versus logistic regression,” Risks, vol. 7, no. 2, 2019, doi: 10.3390/risks7020070.
[20] Z. Dai, X. Dong, J. Kang, and L. Hong, “Forecasting stock market returns: New technical indicators and two-step
economic constraint method,” North American Journal of Economics and Finance, vol. 53, p. 101216, 2020, doi:
10.1016/j.najef.2020.101216.
[21] E. H. Houssein, M. Dirar, L. Abualigah, and W. M. Mohamed, “An efficient equilibrium optimizer with support
vector regression for stock market prediction,” Neural Computing and Applications, vol. 34, no. 4, pp. 3165–3200,
2022, doi: 10.1007/s00521-021-06580-9.
[22] N. Naik and B. R. Mohan, “Optimal feature selection of technical indicator and stock prediction using machine
learning technique,” in Communications in Computer and Information Science, 2019, vol. 985, pp. 261–268. doi:
10.1007/978-981-13-8300-7 22.
[23] K. K. Yun, S. W. Yoon, and D. Won, “Prediction of stock price direction using a hybrid GA-XGBoost algorithm
with a three-stage feature engineering process,” Expert Systems with Applications, vol. 186, p. 115716, 2021, doi:
10.1016/j.eswa.2021.115716.
[24] H. Gunduz, “An efficient stock market prediction model using hybrid feature reduction method based on varia-
tional autoencoders and recursive feature elimination,” Financial Innovation, vol. 7, no. 1, pp. 1–24, 2021, doi:
10.1186/s40854-021-00243-3.
[25] Y. Peng, P. H. M. Albuquerque, H. Kimura, and C. A. P. B. Saavedra, “Feature selection and deep neural networks
for stock price direction forecasting using technical analysis indicators,” Machine Learning with Applications, vol. 5,
p. 100060, 2021, doi: 10.1016/j.mlwa.2021.100060.
[26] Q. Li, G. Yan, and C. Yu, “A novel multi-factor three-step feature selection and deep learning framework for regional
GDP prediction: evidence from China,” Sustainability (Switzerland), vol. 14, no. 8, 2022, doi: 10.3390/su14084408.
[27] B. Xue, M. Zhang, W. N. Browne, and X. Yao, “A survey on evolutionary computation approaches to feature selec-
tion,” IEEE Transactions on Evolutionary Computation, vol. 20, no. 4, 2016, doi: 10.1109/TEVC.2015.2504420.
[28] C. Zhang, Y. Xue, F. Neri, X. Cai, and A. Slowik, “Multi-objective self-adaptive particle swarm optimization for
large-scale feature selection in classification,” International Journal of Neural Systems, vol. 34, no. 3, pp. 1–27,
2024, doi: 10.1142/S012906572450014X.
[29] V. Bolón-Canedo, N. Sánchez-Maroño, and A. Alonso-Betanzos, “A review of feature selection methods on synthetic
data,” Knowledge and Information Systems, vol. 34, no. 3, pp. 483–519, 2013, doi: 10.1007/s10115-012-0487-8.
[30] Y. Chen, H. Jiang, C. Li, X. Jia, and P. Ghamisi, “Deep feature extraction and classification of hyperspectral images
based on convolutional neural networks,” IEEE Transactions on Geoscience and Remote Sensing, vol. 54, no. 10, pp.
6232–6251, 2016, doi: 10.1109/TGRS.2016.2584107.
[31] H. I. Fawaz, G. Forestier, J. Weber, L. Idoumghar, and P. A. Muller, “Deep learning for time series classification: a

Int J Elec & Comp Eng, Vol. 15, No. 2, April 2025: 2192-2201
Int J Elec & Comp Eng ISSN: 2088-8708 ❒ 2201

review,” Data Mining and Knowledge Discovery, vol. 33, pp. 917–963, 2019, doi: 10.1007/s10618-019-00619-1.
[32] X. Li and P. Wu, “Stock price prediction incorporating market style clustering,” Cognitive Computation, vol. 14, no.
1, pp. 149–166, 2022, doi: 10.1007/s12559-021-09820-1.
[33] J. Chun, J. Ahn, Y. Kim, and S. Lee, “Using deep learning to develop a stock price prediction model based
on individual investor emotions,” Journal of Behavioral Finance, vol. 22, no. 4, pp. 480–489, 2021, doi:
10.1080/15427560.2020.1821686.
[34] S. Kumar Chandar, “Grey wolf optimization-Elman neural network model for stock price prediction,” Soft Computing,
vol. 25, no. 1, pp. 649–658, 2021, doi: 10.1007/s00500-020-05174-2.
[35] B. Charbuty and A. Abdulazeez, “Classification based on decision tree algorithm for machine learning,” Journal of
Applied Science and Technology Trends, vol. 2, no. 1, pp. 20–28, 2021, doi: 10.38094/jastt20165.
[36] J. Xianya, H. Mo, and L. Haifeng, “Stock classification prediction based on spark,” Procedia Computer Science, vol.
162, pp. 243–250, 2019, doi: 10.1016/j.procs.2019.11.281.
[37] A. Safari and A. A. Ghavifekr, “International stock index prediction using artificial neural network (ANN) and Python
programming,” in 2021 7th International Conference on Control, Instrumentation and Automation, 2021, pp. 1–7. doi:
10.1109/ICCIA52082.2021.9403580.
[38] D. Sharma, S. K. Singla, and A. K. Sohal, “Stock market prediction using ARIMA, ANN and SVR,” in Lecture Notes
in Electrical Engineering, vol. 698, Springer, 2021, pp. 1081–1092. doi: 10.1007/978-981-15-7961-5 100.
[39] Z. Peng, Q. Huang, and Y. Han, “Model research on forecast of second-hand house price in chengdu based on
XGboost algorithm,” in 2019 IEEE 11th International Conference on Advanced Infocomm Technology, pp. 168–172,
2019, doi: 10.1109/ICAIT.2019.8935894.
[40] Y. Yang, Y. Wu, P. Wang, and X. Jiali, “Stock price prediction based on XGBoost and LightGBM,” in E3S Web of
Conferences, 2021, vol. 275, p. 1040. doi: 10.1051/e3sconf/202127501040.
[41] P. R. Srivastava, Z. Zhang, and P. Eachempati, “Deep neural network and time series approach for finance systems:
predicting the movement of the Indian stock market,” Journal of Organizational and End User Computing, vol. 33,
no. 5, pp. 1–24, 2021, doi: 10.4018/JOEUC.20210901.oa10.
[42] A. Thakkar and K. Chaudhari, “A comprehensive survey on deep neural networks for stock market: The need, chal-
lenges, and future directions,” Expert Systems with Applications, vol. 177, 2021, doi: 10.1016/j.eswa.2021.114800.
[43] W. Lu, J. Li, J. Wang, and L. Qin, “A CNN-BiLSTM-AM method for stock price prediction,” Neural Computing and
Applications, vol. 33, no. 10, pp. 4741–4753, 2021, doi: 10.1007/s00521-020-05532-z.

BIOGRAPHIES OF AUTHORS
Sneha S. Bagalkot currently works as assistant professor in the Department of Com-
puter Science and Engineering, B.M.S. College of Engineering, Bengaluru and is pursuing Ph.D.
in VTU, Belagavi in the Research Center, Department of Computer Science and Engineering, Na-
garjuna College of Engineering and Technology, Bangalore, in the machine learning and data an-
alytics domain. She has about 12 years of teaching experience. She can be contacted at email:
[email protected].

Dinesha H. A. is a serving as professor and dean (RD), at Shridevi Institute of Engi-

neering and Technology. He is a fledgling idealistic entrepreneur indulged passionately in serving
and uplifting society through advance research, innovation, and knowledge imparting in the arena of
engineering and technology. Email id: [email protected].

Nagaraj Naik earned his B.E. degree in Computer Science Engineering from VTU, In-
dia, in 2006, and his M.Tech. degree in 2012. He completed his Ph.D. in 2021 in the Department
of Information Technology at the National Institute of Technology Karnataka, India. His research
focuses on financial data analysis, stock price analysis, and statistical techniques. Email id: na-
[email protected].

A novel technique for selecting financial parameters and technical indicators to ... (Sneha S. Bagalkot)