Insights of effectivity analysis of learning-based approaches towards software defect prediction

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 14, No. 2, April 2024, pp. 1916~1927
ISSN: 2088-8708, DOI: 10.11591/ijece.v14i2.pp1916-1927  1916
Journal homepage: http://ijece.iaescore.com
Insights of effectivity analysis of learning-based approaches
towards software defect prediction
Deepti Rai, Jyothi Arcot Prashant
Department of Computer Science and Engineering, M. S. Ramaiah University of Applied Sciences, Bengaluru, India
Article Info ABSTRACT
Article history:
Received May 30, 2023
Revised Jul 12, 2023
Accepted Dec 13, 2023
Software defect prediction is one of the essential sets of operation towards
mitigating issues of risk management in software development known to
contribute towards enhancing the quality of software. There is evolution of
various methodologies towards resolving this issue while learning-based
methodology is witnessed to be the most dominant contributor. The problem
identified is that there are yet many unsolved queries associated with
practical viability of such learning-based approach adoption in software
quality management. Proposed approaches discussed in this paper
contributes towards mitigating this challenge by introducing a simplified,
compact, and crisp analysis of effectiveness associated with learning-based
schemes. The paper presents its major findings of effectivity analysis of
machine learning, deep learning, hybrid, and other miscellaneous approaches
deployed for fault prediction followed by highlighting research trend. The
major findings infer that feature selection, data imbalance, interpretability,
and in adequate involvement of context are prime gaps in existing methods.
The paper also contributes towards research gap as well as essential learning
outcomes of present review work.
Keywords:
Deep learning
Machine learning
Software defect prediction
Software fault
Software quality
This is an open access article under the CC BY-SA license.
Corresponding Author:
Deepti Rai
Department of Computer Science and Engineering, M. S. Ramaiah University of Applied Sciences
New BEL Rd, M S R Nagar, Mathikere, Bengaluru, Karnataka 560054, India
Email: deeraisecond@gmail.com
1. INTRODUCTION
The organization involved in development of software products are required to deal with various
risk and threats in its development process irrespective of the size of an organization and one such important
risk factor is associated with software defects [1]. There are various types of software defects viz.
configuration defect, maintainability defects, compatibility defect, usability defect, security defect,
performance defect, and functional defect [2]–[8]. The presence of defect in the finally developed software
results in higher amount of risk associated with financial and resources which can be controlled when an
effective defect predictive is applied. The mechanism of software defect prediction involves using statistical
models as well as software metric in order to determine the potential defect before even they surface in real
time [9]. The prim intention of software defect prediction is mainly to enhance the quality of software as well
as minimize the software development cost by early evaluation of risk in the form of software defect [10]. At
present, there are various standard methodologies and approaches associated with software defect prediction
viz. i) the first type of such approach is termed as static-analysis based predictive approach where the source
code is evaluated and critical defects are identified on the basis of security vulnerabilities, inconsistency, and
coding errors [11], ii) the second type of such approach is called as dynamic-analysis based predictive
approach which is carried out during runtime of software by monitoring its behavior in the form of different

Int J Elec & Comp Eng ISSN: 2088-8708 
Insights of effectivity analysis of learning-based approaches towards … (Deepti Rai)
1917
runtime errors, buffer overflows, and memory leaks [12], iii) the third type is known as model-based
predictive approach which make use of either machine learning or statistical model in order to perform
prediction. Usually, such process depends upon the initial training with the standard dataset which in
facilitated with both information about defects and software quality [13], iv) the fourth type is called expert-
based predictive approach which is completely based on experience and skilled knowledge of domain expert
in order to determine the defect [14], and v) the last type of approach is called as hybrid-based predictive
approach which integrates multiple variants of above-mentioned approaches for similar cause [15]. Out of all
this, machine learning approaches are found to be frequently adopted owing to their potential ability to learn
the unique patterns from the massive-sized dataset of software metric in order to perform prediction. Apart
from this, it is also found that such approaches also consider usage of specific form of metrics viz.
i) cohesion-based metric (quantity of methods in specific module or class), ii) coupling metric (quantity of
inter-module dependencies), and iii) complexity-based metric (quantity of conditionals, depth of nesting, size
of code) [16]. Adoption of such metric is carried out in order to perform machine learning-based training
with an agenda of forecasting the presence of defects in specific modules or classes. After the training is
accomplished, different classes or modules with maximal risk probability is subjected for identification.
Prior to understanding and realize the statement of the problem, it is essential to determine the
varied labels of issues and ongoing challenges associated with software defect prediction as follows: i) the
first research problem is associated with establishing the connectivity between the defect and the attributes,
which is one of the significant factors for ascertaining the accuracy of defect. Further, adoption of accurate
metric level is still undefined; although there are various studies being carried out; but the outcome have been
accomplished over different set of attributes, where a proper inference cannot be drawn effectively, ii) there
is lack of robust and reliable parameter towards assessing the performance of such predictive tool; apart from
this adoption of standard criterion is significantly missing in existing scheme which can be proven for wide-
range of applicability in multiple and different software projects, iii) usage of local data is carried out towards
model learning from some old project; however very often the development team encounters higher
dependencies of such local data which are unavailable sometimes due to manifold reason. Prediction scheme
using cross-projects is one of the alternative solutions to deal with this issue where prediction of defect is
carried out in specific project while analysis is carried out in different software project. Unfortunately,
predictive scheme developed on cross-project may have claimed for higher accuracy; but their outcome is
strictly restricted to some set of software domain and are sub-optimal in its predictive analysis, iv) at present,
there is no report of any available generalized framework for software defect prediction; adoption of different
predictive methodologies on different types of dataset is the prime reason for this issue, v) majority of the
research models towards software defect prediction is carried out on the basis of simplified usage without any
emphasis towards cost modelling associated with various uncertain factors. This results in misclassification
as well as interpretability issues, and vi) class imbalance has always been a critical problem in software
defect prediction irrespective of various literatures being consistently addressing them. Therefore, the clear
statement of the problem can be arrived as “developing an optimal predictive scheme towards accurately
determining software defect with balanced computational efficiency and wide range of applicability on
different software project is yet a computationally challenging task”.
The above-mentioned issue can be more effectively justified on the basis of varied ranges of
relevant literatures. Adoption of machine learning approaches towards prediction of defects explicitly
considering the use case of mobile application is discussed by Jorayeva et al. [17]. The discussion carried out
by Abdu et al. [18] have stated varied methods for usage of semantic attributes using deep learning
approaches for predicting the software defects. Further work stating the significance of deep learning
approaches is noted in work of Akimova et al. [19] and Giray et al. [20]. Aziz et al. [21] has presented
discussion associated with multiple inheritance metrics in order to perform analysis of software fault
prediction. Generalized discussion towards various learning-based methodologies were carried out by
Cao [22], Mahmud et al. [23], Son et al. [24] and Kotte and Qyser [25]. From the perspective of machine
learning approach, Khan et al. [26] have presented discussion of predictive approaches using artificial neural
network (ANN). Pal and Sillitti [27] have presented discussion about various predictive approaches for cross-
projects. All the above-mentioned relevant literatures proves that there are significant number of research
studies towards varied solution in software defect prediction.
However, it is yet not totally clear to understand the best solution or clear and pin-pointed
interpretation of research gaps. Therefore, the proposed study manuscript presents a compact and yet clear
visualization of the effectiveness of some of the recent and most relevant schemes of software defect
prediction. The core objective of this manuscript is to review the potential approaches reported in current era
towards software defect prediction with crisp information. The new value of the research is presented in the
form of following contribution of this manuscription as i) the study reviews most significant literatures
associated with both learning and non-learning approaches, ii) all the reviewed literatures have been
exhaustively reviewed with respect to problems being addressed, specification of adopted methodologies,

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 14, No. 2, April 2024: 1916-1927
1918
reported beneficial factors, and identified shortcoming, iii) a highlights of recent research trends towards
adoption of individual approaches of both machine learning and deep learning is carried out to understand the
frequency of usage of specific methodologies, iv) a clear insight is furnished towards the identified research
gap from the review work to determine the open-end challenges associated with it, and v) learning outcome
of the study in the form of significant contribution to further offer a researcher’s viewpoint towards
facilitating further direction of study by considering constructive suggestion. Hence, the proposed review
work offers a crisp, updated, and informative contents towards adoption of learning-based approaches in
software defect prediction. The next section outlines the adopted methodology in order to carry out the
present review work.
2. METHOD
The prime purpose of the proposed study is to carry out a review of learning-based approach
towards understanding its strength and weakness towards improving software defect prediction. A desk
research methodology is adopted for this purpose with simplified sequence of following task as shown in
Figure 1 viz: i) the first task is to perform an information identification where the publication of research
articles from reputed archives with high impact factor is seeked. In this phase, all the research articles
published between 2013 to till date has been collected which mainly deals with implementation plans; ii) the
second task is to perform initial filtering of all the collected articles by screen the abstract to ensure that it
meets the agenda of study i.e., to find implementation strategy; iii) the third task is to perform identification
and removal of duplicates which refers to exactly same implementation strategy or different papers written
by same author bearing nearly the same techniques; and iv) the fourth task is to perform a secondary
screening where the methodology and result is screened. The methodology is screened to understand the
uniqueness of the algorithm and technique used for software fault prediction while result screening is to find
the strength and weakness of the associated implemented methodology. This task of both primary and
secondary screening is carried out considering inclusion and exclusion criteria.
The inclusion criteria involve only the research article bearing implementation plan with enough
evidence of result work while the exclusion criteria involve considering any implementation paper published
before 2013. Finally, the outcome of the secondary screening process results in extraction of research gap as
well as extraction of learning outcomes. The research gap is extracted on the basis of review of limitation
extended to further understanding the global perspective of the methodology under screening. The learning
outcomes consists of researcher’s self-opinion which is based on perspective of complete review of strength
and weakness of existing methodologies. In the entire course of this methodology, equal extraction process is
carried out towards understanding the research trend on the basis of various phases of methodology involved
in proposed study. The idea of research trend is to look for frequency of adoption of different learning-based
scheme towards software defect prediction. Therefore, this methodology, in its simplified form, assists in
accomplishing research objectives.
Figure 1. Adopted method in present investigation
Information
Identification
Initial Filtering
Identifying &
removing
duplicates
Abstract Screening
Secondary
Screening
Methods & Result
Screening
Inclusion
Criteria
Exclusion
Criteria
Extract Research Gap
Extract Learning
Outcomes
Screen
Research
Trend

1919
3. RESULTS
This section presents the briefing of the study findings associated with the usage of different
approaches towards prediction of software fault. There are different variants of techniques adopted towards
this purpose where machine learning as well as deep learning-based algorithms are dominant techniques
while some other approaches are reviewed as well. The core agenda of the discussion of this section is to
highlight the effectiveness of all the reviewed research articles to understand their further applicability to deal
with the problems of software fault prediction.
3.1. Existing studies deploying machine learning approaches
Adoption of machine learning-based approaches has been witnessed in different way in order to
address different challenges associated with software fault prediction. The work carried out by Aftab et al.
[28] have used a combination of decision tree (DT), artificial neural network (ANN), and naïve Bayes (NB)
in order to carry out classification of different faults in software design on cloud. The system also used fuzzy
logic (FL) in order to incorporate accuracy associated with the prediction. Problems associated with the sub-
optimal solution due to usage of different ensemble model for prediction of software fault is addressed in
work of Alazba and Aljamaan [29] where hyperparameters have been tuned up for multiple ensemble
approaches with tree design e.g., CatBoost, XGBoost, gradient boosting using histogram, normal gradient
boosting, AdaBoost, and random forest. Adoption of ensemble approach towards classification problem was
witnessed in work of Alsawalqah et al. [30] where a simplified classifier and ensembled classifier is designed
for robust classification. Aziz et al. [31] have used standard software metric i.e., Chidamber and Kemerer
(CK) metric along with machine learning for investigating the possible influence of this metric on predicting
software fault. The study uses ANN for model building considering repositories with and without inheritance
of CK metric. Bal and Kumar [32] have addressed the issue of data imbalance while perform prediction of
software fault by introducing a weighted regularization scheme of machine learning approach in order to
accomplish a balanced outcome. Study towards similar problem of data imbalance is also carried out by
Khout and Le [33] where ensemble learning approach is used for prediction of software fault. The study
model uses DT, support vector machine (SVM), k-nearest neighbor (KNN), Bayesian network (BN), and
multilayer perceptron (MLP) for performing classification.
Some of the studies have linked software defects with hardware failures in existing era and
introduces a solution for this as reported in work of Boateng et al. [34]. According to this study model,
feedforward neural network and linear regression method has been used for investigating the cost associated
with hardware failures for optic networks. Adoption of machine learning has been further investigated by
Khoudry et al. [35] using K-nearest neighbor and Gaussian process towards identifying fault associated with
high-speed power mechanism. The work carried out by Lee and Seo [36] have used KNN and SVM along
with latent Dirichlet allocation (LDA) in order to enhance the software bug reporting system. Adoption of
unsupervised learning scheme is witnessed in work of Marjuni et al. [37] by using spectral classifier to deal
with zero-threshold problems. The study presents a unique threshold with median absolute deviation
considering eigen vector. Further, a unique case study is considered by Nasir et al. [38] in order to identify
the level of tolerance that can offer by information-centric software. Voutsinas et al. [39] have used a
machine learning approach in order to identify the fault considering photovoltaic system. The idea of this
study model is to perform classification of different types of faults associated with use-case while it also
reduces the computational cost associated with its operation in terms of minimal memory.
3.2. Existing studies deploying deep learning approaches
Deep learning approaches has been recently evolved to prove its effectiveness towards prediction of
software fault owing to its potential advantage towards accuracy accomplishment in its analytical process.
The work carried out by Deng and Qiu [40] have addressed the problems of semantic characteristic
associated with programming language towards generation of defective code by using long short-term
memory (LSTM). This is used for learning the contextual and semantic features associated with the source
code. The mechanism calls for constructing an abstract tree of the program in order to evaluate each data
within the tree nodes. Adoption of LSTM was also reported in work of Munir et al. [41] where it has been
integrated with gated recurrent unit (GRU) with an objective of classifying faults. The source code is
subjected to parsing and a tree is formulated along with incorporation of 32-level matrix features.
The work carried out by Hai et al. [42] have presented a technique to track the bugs present within
cloud environment using deep learning. The core notion of this study model is to minimize the cost and time
demands for performing assessing of software defect tracking. The scheme has been implemented using
multiple multi-layer perceptron (MLP) configuration over multiple standard datasets. Adoption of deep
learning is also witnessed in work of Jorayeva et al. [43] where LSTM is integrated with convolution neural
network considering the user case of open-source Android application software defects. The analysis also has
perform comparative analysis for ANN, convolution neural network (CNN), and LSTM to show that

 ISSN: 2088-8708
1920
performance of CNN and LSTM is always better than ANN while CNN is slightly more better than LSTM.
Similar adoption of CNN and LSTM integration is also reported in work of Farid et al. [44]; however, the
authors have used bidirectional LSTM (Bi-LSTM) towards prediction of software defect caused due to
semantics associated with the source code. The study model formulates a syntax tree to represent the vectors
linked with programs where the extraction of semantics is carried out by CNN model and retention of key
features are carried out by Bi-LSTM. A unique form of deep learning approach known as contrastive learning
is implemented by Luo et al. [45] towards enhancing the identification of software defect in the form of bug.
The study model initially performs pretraining on the corpus of bug reports using unsupervised scheme of
learning followed by training with contrastive learning. The contribution of this scheme is that it assists in
learning the semantic distinction between buggy files and defect reports. Another unique adoption of deep
learning is carried out by Maduako et al. [46] where the defect analysis is carried out for power transmission
lines. This scheme re-tunes the CNN model along with a novel pyramid network with multiscale layer in
order to localize the faults. Pan et al. [47] have developed a computational model where an enhanced CNN
model is used for predicting software faults. The author has developed a new dataset of source codes where
CNN has been implemented and the outcome shows that improved CNN offers better performance on defect
prediction compared to conventional CNN. Qasem et al. [48] have investigated the impact of deep learning
towards fault prediction by integrated usage of MLP and CNN. Wang et al. [49] have used gated LSTM of
hierarchical form where the semantic features of the codes are extracted in the form of syntax tree. The
model is claimed to be capable of extracting both conventional and semantic features of software using gated
fusion technique. The further studies towards mining semantic feature have been carried out by
Yao et al. [50]. According to this study model, a CNN model is developed in the form of tree-based
structured where extraction of semantic feature is carried out from grammatical structure of code as well as
text information within the code.
3.3. Existing studies deploying hybrid learning approaches
There are also certain studies where hateful and offensive speech detection is carried out jointly. The
work It is known that adoption of both machine learning as well as deep learning has been frequently
exercised in order to accomplish a better prediction performance for software defect. However, there are also
approaches which has combined both machine learning and deep learning scheme in order to form a hybrid
approach to further harness the predictive potential of both the learning schemes for optimized performance.
One such unique and simple form of a hybrid learning model has been presented in work of Asmawi et al.
[51] by integrating deep learning and machine learning approach. The idea of this work is towards predicting
the failures of cloud-based software. The findings of the study state that extreme gradient boosting is found to
be suitable model towards processing essential features associated with disk and central processing unit
(CPU) spaces while random forest and DT method is found be suitable model towards task prioritization in
the course of software defect prediction. Borandag et al. [52] have developed another hybrid scheme by
integrating recurrent neural network (RNN) with ensemble machine learning technique. The authors have
used CNN, LSTM, and Bi-LSTM as a part of deep learning approach while 5 machine learning techniques
(i.e., naïve Bayes, SVM, KNN, random tree, and K-Star) over benchmarked dataset evaluated over
object-oriented metric. Khalid et al. [53] have presented another unique model towards software defect
prediction where machine learning technique has been integrated with bio-inspired algorithm in order to
formulate a new hybrid learning algorithm. The study has implemented K-means clustering as machine
learning approach integrated with particle swarm optimization (PSO) as a part of bio-inspired algorithm in
order to perform classification of features. The study outcome shows that SVM offers the best optimal results
compared to other machine learning approach. Similar trend of hybrid approach is also carried out by
Zhang et al. [54] where a back propagation neural network is integrated with cuckoo search optimization
scheme. The idea of this study model is to perform prediction of faults in industrial equipment’s with an
outcome exhibiting improved training time and convergence performance.
3.4. Other approaches
Apart from conventional learning-based algorithms, there are various approaches towards
performing prediction of software defects. It has been noted that adoption of cross-projects analysis of defect
is one of the better alternative solutions to overcome the issues of historical records; however, it is also
characterized with shortcoming associated with prime distinction between target software project with source
project. This issue is addressed in work of Bala et al. [55] who have adopted a selection scheme on the basis
of features and transformation. Adoption of transformation method is towards minimization of difference in
distribution of projects while minimization of high dimensionality and unnecessary attributes are governed
by selection of optimal features in this study model. Further, the study model presented by
Hassouneh et al. [56] have used bio-inspired approach for addressing the challenges in software defect

1921
prediction. The authors have implemented whale optimization algorithm where the exploration process
towards optimal outcome is improved. Further, process improvement is carried out using multiple method of
selection i.e., random, stochastic universal sampling, linear rank, roulette wheel, and tournament. A distinct
scheme towards prediction of software fault is discussed by Lee et al. [57] where an analytical model is
presented in the form of a software network that connects both software module and developers in order to
investigate the interaction between them. The idea of this concept is to derive the sub-graph that is
characterized by bad structure of research objects in order to indicate the software fault. Study towards
similar form of concept is implemented by Li et al. [58] where a tri-relation network is designed for
prediction of software faults. This form of network integrates developer, module dependency, and
contribution of developer in order to investigated their joint influence towards software quality. The work
carried out by Phung et al. [59] have presented a unique scheme where software metric is formulated for
error representation associated with java-based runtime. A formal modelling method is used along with
machine learning in order to evaluate the patterns. The work of Tumar et al. [60] have used moth flame
optimization scheme in order to carry out selection of feature associated with fault in software design. The
study has also used an adaptive synthetic sampling method along with moth flame optimization in order to
perform selection of wrapper feature as well as to mitigate the issue of imbalanced dataset. The uniqueness of
this study is also to convert the bioinspired approach to a binary version using transfer function. Further, the
study model has also used linear discriminant analysis (LDA) along with DT and KNN for performing
classification. The work carried out by Zhang et al. [61] have developed a scheme to predict the defect
associated with labelled data using cross-version model. The implementation design contributes towards
solving issues related to class overlapping and distribution of data variation.
Table 1 highlights the summarization of the all the reviewed existing scheme deploying machine
learning, deep learning, hybrid method, and other miscellaneous unique methodology towards software fault
prediction. The tabular content exhibits that the existing system has adopted different variants of techniques
which are associated with claimed beneficial features as well as they are also characterized with limiting
features. The inference of the learning outcome of this table is that although existing software fault prediction
has made some significant progress but still there are associated limitation associated with almost all the
adopted methodologies.
3.5. Research trend
At present there are various research work being carried out towards software fault prediction
mainly using machine learning and deep learning approaches in last decade. According to the information
stated in Table 2, it can be noted that there are approximately 74,032 research articles published in last ten
years. It should be noted that these publications are cumulative of all learning and non-learning-based scheme
towards enhancing the predictive performance of software fault detection. Another indication inference from
Table 2 is that there is a significant number of interests being shown from the research community towards
software fault prediction owing to increasing number of research articles.
However, it is further essential to know the contribution of core learning-based approaches in this
regard. Therefore, this research trend is exhibited in Table 3 and Table 4 where contribution of machine
learning and deep learning approaches are exhibited. Table 3 has exhibited deployment of various frequently
used machine learning schemes viz. Naïve Bayes (NB), decision tree (DT), support vector machine (SVM),
random forest (RF), k-nearest neighbor (KNN). The outcome of this research trend showcases that adoption
of SVM is consistently on rise followed by adoption of KNN and NB approach. The deployment of DT and
RF is found to be less adopted with progression of research publications. One interesting point to be noted
here is that in an era of technological advancement in artificial intelligence, the adoption of conventional
machine learning approach SVM, KNN, and NB is still in use owing to its potential advantage towards
predictive performance in software fault detection. It is also noted that these techniques are adopted in two
ways viz. i) acts as a core model for prediction or ii) it acts as a benchmark model for comparative assessment
of presented predictive model. In either of both the cases, they are highly helpful from research perspective.
Table 4 highlights the trends of adoption of deep learning methods viz. Convolution neural network
(CNN), long short-term memory (LSTM), recurrent neural network (RNN), generative adaptive network
(GAN), radial basis function network (RBFN), multilayer perceptron (MLP), and self-organizing map
(SOM). It is noted that LSTM is one of the deep learning approaches that has been extensively deployed
which is followed by equal number of usages of CNN and GAN. Although, RNN has been also equivalently
deployed; however, majority of the research article has reportedly used RNN as initial process followed by
finally deploying its enhanced variant i.e., LSTM. Adoption of SOM is still on steady pace but not
extensively found to be used as compared to other variants of deep learning approaches. Apart from this, the
analysis towards non-learning approaches is really scattered and no significant trend is noticed and still it
forms a minority-approaches towards software defect prediction.

 ISSN: 2088-8708
1922
Table 1. Summary of existing approaches
Authors Problem Methodology Advantage Limitation
Aftab et al.
[28]
Software defect prediction DT, ANN, NB, FL 91.05% accuracy Cost effectiveness not
analyzed
Alazba and
Aljamaan [29]
Usage of default
hyperparameters in
ensemble
Optimization of
hyperparameter of multiple
ensemble
Effective optimization Model does not have
practical constraints
Alsawalqah
et al. [30]
Software defect prediction Hybrid ensemble Supports heterogeneous
classification
Not applicable for dynamic
faults
Aziz et al. [31] Impact of CK metric on
Fault prediction
ANN Proves the importance of
inheritance in prediction
Not benchmarked with other
prediction approach
Bal and Kumar
[32]
Data imbalance in
prediction
Weighted regularization,
feed forward neural network
Effective fault detection Study does not perform
binary classification
Khuat and Le
[33]
Data imbalance in
prediction
DT, SVM, KNN, BN, and
MLP
Simplified model No constraints in modelling
Nyarko-
Boateng [34]
Cost evaluation in
hardware failures
Linear regression,
feedforward neural network
Satisfactory accuracy,
practical utilization
No benchmarking
Khoudry et al.
[35]
Fault prediction for high-
speed power system
KNN, Gaussian Works both online and
offline
Dynamic faults are not
associated with the model
Lee and Seo
[36]
Improving software bug
reporting system
LDA, KNN, SVM Increased accuracy Use-case specific
performance
Marjuni et al.
[37]
Zero threshold issues Unsupervised Learning,
heuristic row sum method
Improved performance of
classification
No benchmarking
Voutsinas
et al. [39]
Fault detection in
photovoltaic system
Machine learning Simplified learning
system, 97.11% accuracy
Needs further extensive
analysis, specific to use case
Deng et al.
[40]
Contextual/semantic
feature challenges
LSTM Simplified and user-
friendly model
Less extensive analysis
Munir et al.
[41]
Software defect prediction LSTM Effective classification
performance
No benchmarking
Hai et al. [42] Software defect tracking
in cloud
MLP Highly simplified model
implementation
Demands higher iteration
and training dependencies
Jorayeva et al.
[43]
Software defect prediction
in mobile application
LSTM, CNN 93% accuracy Dynamic defects not
assessed
Farid et al.
[44]
Semantics linked with
source codes
Bi-LSTM, CNN Improve detection of
faults
Assessed only on java
projects
Luo et al. [45] Localization of bug Contrastive Learning Complete automated
contextual model
Models demands predefined
information of bugs
Maduako et al.
[46]
Component fault detection CNN, pyramid network with
multiscale feature
Supports representation of
multiple fault
characteristic
Study model specific to data
type
Pan et al. [47] Software defect prediction Improved CNN Better detection
performance
Computational complexity
issue
Qasem et al.
[48]
Software defect prediction CNN, MLP Comprehensive analytical
model
Induces computational
burden
Wang et al.
[49]
Software defect prediction LSTM (gated) Distributes the
computational
complexities during
prediction
No benchmarking
Yao et al. [50] Semantic Extraction CNN, feature mining of
semantics
Higher performance score Lower assessment scope
Asmawi et al.
[51]
Prediction of cloud-based
defects
Hybrid learning Short time for prediction Low accuracy score
Borandag [52] Software defect prediction Hybrid learning 95% accuracy Demands more processing
resources
Khalid et al.
[53]
Software defect prediction Machine learning,
bioinspired algorithm
Optimal feature learning Premature convergence not
addressed
Zhang et al.
[54]
Fault prediction in
industrial equipment
Cuckoo search optimization,
backpropagation neural
network
Good convergence
performance, and reduced
training time
Model not applicable for
dynamic/uncertain defects
Bala et al. [55] Cross project software
defect prediction
Transformation & selection
of feature
Can process high
dimensional feature
Scalability issues
Hassouneh
et al. [56]
Feature selection in
software defect prediction
Whale optimization High data quality, higher
reliability
No consideration of
uncertain risk attributes
Lee et al. [57] Human-based software
errors
Tree network of developer
and module
Can be customized based
on demand
Highly sensitive to slightest
fault
Li et al. [58] Human-based software
errors
Tri-relational network Simplify the process of
prediction
Needs extensive analysis to
proves it applicability
Phung et al.
[59]
Fault identification Software metric with error
type, formal modelling,
machine learning
Can identify run-time
error
Applicable for java
environment only
Tumar et al.
[60]
Feature selection for fault
prediction
Moth flame optimization,
machine learning, transfer
function
Improved classification
performance
Iterative model
Zhang et al.
[61]
Challenges in fault
prediction in labelled data
Cross-version model Effective clustering
performance
Computational complexity
not assessed

1923
Table 5 highlights trends towards the adoption of different dataset [62]–[80] which are reported to
be frequently deployed in existing studies. However, a unique trend is noted towards the usage of such
standard and frequently reported dataset. For an example, it is noted that adoption of University College
London (UCL) machine learning (ML) repository is used for varied forms of machine learning approaches
itself while adoption of NASA, PROMIZE, and GHPR dataset is reported to be used more using AdaBoost
and bagging, Naïve Bayes, random forest [79]. It is also noted that adoption of Poi, Synapse, Xerces, Xalam,
Lucene, jEdit, Camel dataset is used for CNN-based approaches mainly. The dataset of AR6, AR4, AR5,
jEdit 4.3, AR3, jEdit 4.2, AR1, jEdit 4.0, Anr 1.7, Tomcat 6.0 has been reportedly used for deploying
HyGRAR modelling. The dataset KC1, KC3, PC5, PC3, PC4, MC2, PC1, MC1, CM1, CM1, MW1, JM1 is
mainly used towards study model using SVM, MLP, D, KNN, RF, [80]. All these datasets consist of
approximately 22-40 attributes (class attribute, size attribute, McCabe attribute, Halstead attribute) with
number of instances ranging between 63-9593 while all the dataset bears a numerical data type.
Table 2. Trend of research publication in software fault prediction
Publication No.
IEEE 144
MDPI 54
Springer 12863
Elsevier 22594
Hindawi 5278
ACM 27500
Taylor and Francis 5599
Table 3. Trend of usage of machine learning approaches in software fault prediction
Publication NB DT SVM RF KNN
IEEE 2 6 10 7 3
MDPI 3 2 6 2 1
Springer 2 15 7 7 4
Elsevier 898 4558 6464 2905 2427
Hindawi 6 34 48 30 15
ACM 30801 32679 34520 32433 34032
Taylor & Francis 158 908 855 437 699
Table 4. Trend of usage of deep learning approaches in software fault prediction
Publications CNN LSTM RNN GAN RBFN MLP SOM
IEEE 6 3 4 2 0 2 5
MDPI 0 2 1 0 0 0 0
Springer 1 2 1 1 0 0 8
Elsevier 2835 2429 2017 571 1751 853 910
Hindawi 41 27 7 10 0 8 5
ACM 31507 34660 31588 33941 34384 27643 33817
Taylor & Francis 188 407 169 80 338 105 714
Table 5. Trend of usage of frequently used dataset for software fault prediction
Publication Work Carried out by
UCI ML repository Khan et al. [26]
NASA Dataset Bowes et al. [62], Pandey et al. [63], Chen and Dai [64],
Mustaqeem and Saqib [65], Marapelli et al. [66]
Poi, Synapse, Xerces, Xalam, Lucene, jEdit, Camel Dam et al. [67], Farid et al. [44], Hosseini et al. [68],
Sikic et al. [69], Li et al. [70]
AR6, AR4, AR5, jEdit 4.3, AR3, jEdit 4.2, AR1, jEdit 4.0,
Anr 1.7, Tomcat 6.0
Miholca et al. [71]
KC1, KC3, PC5, PC3, PC4, MC2, PC1, MC1, CM1, CM1,
MW1, JM1
Iqbal et al. [72]
PROMISE dataset Pachouly et al. [73], Zain et al. [74], Bal et al. [75], Bahaa et al. [76]
GHPR dataset Batool and Khan [77], Pan et al. [78]
Therefore, from the highlight of information discussion in this section in perspective of research
trend, it is noted that there is a unique pattern of deployment of learning-based approaches in software defect
prediction, where a greater number of research articles is witnessed for deep learning-based approach
(= 2,41,043) compared to machine learning approach (n=1,84,984) in last ten years. Apart from this, it is also

 ISSN: 2088-8708
1924
noted that various dataset too has unique mechanism of adoption which is specific to learning-based
approach. However, more adoption of standard dataset is reported for machine learning schemes in contrast
to deep learning schemes towards software defect prediction. The next section discusses about the learning
outcomes of the review with respect to research gap.
3.6. Research gap
After reviewing the existing implementation approaches of different variants towards software
defect prediction, it is noted that that both machine learning and deep learning are the most dominant
approaches owing to its beneficial features. However, there are shortcomings associated with both the
approaches which is reported in prior section. A closer insight to shortcomings of both the learning
approaches also exhibit that there are various critical open-end problems that are yet not reported to be
addressed in existing research models. Following are the research gap explored:
a. Inability to furnish proper contextual information: majority of experiments of the existing learning-based
models towards software defect prediction is carried out over historical data in order to offer a predictive
outcome. However, the predictive data lacks any form of consideration towards contextual attributes e.g.,
requirements of software project, or any alterations in personnel or development practices.
b. Issues of interpretability: from the viewpoint of application-based interpretation, it is noted that machine
learning offers more user-friendly interpretability compared to deep learning approaches. Adoption of
deep learning approaches significantly increases accuracy levels but cannot provide clear interpretability
of its outcome with respect to the associated defects.
c. Unclear evidence of applicability: adoption of both machine learning and deep learning were proven to
accomplish its claimed results in existing studies addressing the specific set of problem. However, it is to
be noted that such accuracy level or betterment in its outcome is only applicable if it is used with same
training data. It is still unclear about its applicability in practical environment where test data could be
highly dynamic and challenging even to process and thereby impose issues in predictability as well.
d. Challenges in feature selection: from the feature engineering viewpoint, it is known that deep learning offers
better outcome compared to machine learning approach. However, these hypothetical assumptions vary
from case to case. At present, different variants of work carried out does not offer a concrete justification
behind the selection of feature with the success factor of the model to address the problem as much.
e. Issues of data imbalance: existing studies adopts different available standards datasets for predicting
software defects; however, in real environment, the availability of information towards software detection
could be very vague or less information. This adoption often leads to biased model in practical application
within an organization where massive number of software projects with different challenges in
information is required to be analyzed.
f. Challenges in integration: different levels of software development team use different standards for their
software project development by adopting different methodologies. However, the adoption of
methodologies for project development and identification of software bugs are always considered as two
different tasks. Integrating development environment with defect detection is still an open-end issue that
further requires a solution.
g. Challenges in ascertaining suitable dataset: different researchers adopt different dataset in their
experiments to prove the efficiency of their study model. However, the data in real words potentially
lacks any form of labelling in its information. Hence, acquiring information and labelling them is quite a
computationally extensive task yet to meet its effective solution.
4. CONCLUSION
Predicting software defect has been investigated since more than a decade and has evolved with
various solution; however, adoption of learning-based solution is quite high. The paper has offered discussion
about exclusively about the usage of machine learning as well as deep learning-based methodologies towards
software defect prediction as well as it has also highlighted usage of hybrid models and other miscellaneous
methodologies reported. The study evaluation found advantages as well as shortcoming associated with each
and every technique while the proposed review contributes towards highlighting the essential research gap.
Further after reviewing the overall study approaches, various learning outcomes in the form of review
contribution have been arrived as: i) The existing problems towards selection of an optimal features while
applying learning approach is to adopt potential feature engineering method towards enhancing accuracy.
Further, an integrated adoption of feature selection along with scaling of feature and reduction of
dimensionality problems will be further more beneficial towards reducing computational burden of learning-
based approaches; ii) Adoption of hybrid approaches as well as ensemble approaches are less in number in
existing studies and hence, it is advisable towards more adoption of hybrid scheme towards minimization of

1925
model variance that can significantly increase the ability of generalization of learning model. Various
techniques e.g., stacking, boosting, and bagging. should be involved in order to address this issue;
iii) Although, transfer learning is an integral part of CNN, but it could be used individually too as a distinct
set of operators that can actually improve learning performance on different set of tasks. The ability of
gaining knowledge can be significantly improved upon usage of transfer learning especially in presence of
low scaled data availability; and iv) Adoption of active learning is also noticed to be less incorporated in
existing schemes whereas this form of learning offers opting for highly informative data to be labelled by an
expert. This adoption will further minimize the computational effort towards labelling demands and directly
enhance the predictive accuracy. This form of learning is one of the better alternatives while dealing with
imbalanced dataset or during any investigation where the labelling cost can be anticipated as maximum.
Therefore, the future direction of the study will be to act upon above-mentioned learning outcomes of this
review study.
REFERENCES
[1] T. O. Olaleye, O. T. Arogundade, S. Misra, A. Abayomi-Alli, and U. Kose, “Predictive analytics and software defect
severity: a systematic review and future directions,” Scientific Programming, vol. 2023, pp. 1–18, Feb. 2023, doi:
10.1155/2023/6221388.
[2] U. Bhardwaj, A. E. Sand, and M. Warrier, “Stability of <100> dislocations formed in W collision cascades,” Journal of Nuclear
Materials, vol. 569, Oct. 2022, doi: 10.1016/j.jnucmat.2022.153938.
[3] M. Maddeh, S. Ayouni, S. Alyahya, and F. Hajjej, “Decision tree-based design defects detection,” IEEE Access, vol. 9,
pp. 71606–71614, 2021, doi: 10.1109/ACCESS.2021.3078724.
[4] J. Liao and W. Feng, “Compatibility defects of the fiber-optic liquid level and refractive index sensors based on modal
interference,” Physica B: Condensed Matter, vol. 624, Jan. 2022, doi: 10.1016/j.physb.2021.413398.
[5] N. S. M. Yusop, J. Grundy, J.-G. Schneider, and R. Vasa, “Preliminary evaluation of a guided usability defect report form,” in
2018 25th Australasian Software Engineering Conference (ASWEC), Nov. 2018, pp. 81–90, doi: 10.1109/ASWEC.2018.00019.
[6] M. A. Puentes, Y. Lei, N. Rakotondravony, L. T. Harrison, and C. A. Shue, “Visualizing web application execution logs to
improve software security defect localization,” in 2022 IEEE International Conference on Software Analysis, Evolution and
Reengineering (SANER), Mar. 2022, pp. 1183–1190, doi: 10.1109/SANER53432.2022.00138.
[7] Y. Zhao, K. Damevski, and H. Chen, “A systematic survey of just-in-time software defect prediction: online supplement,” ACM
Computing Surveys, vol. 55, no. 10, pp. 1–35, Feb. 2023, doi: 10.1145/3567550.
[8] H. Wang and L. Yuan, “Software engineering defect detection and classification system based on artificial intelligence,”
Nonlinear Engineering, vol. 11, no. 1, pp. 380–386, Jan. 2022, doi: 10.1515/nleng-2022-0042.
[9] I. Arora and A. Saha, “ELM and KELM based software defect prediction using feature selection techniques,” Journal of
Information and Optimization Sciences, vol. 40, no. 5, pp. 1025–1045, Jul. 2019, doi: 10.1080/02522667.2019.1637999.
[10] A. Sinha, S. Singh, and D. Kashyap, “Implication of soft computing and machine learning method for software quality, defect and
model prediction,” in Multi-Criteria Decision Models in Software Reliability, CRC Press, 2022, pp. 45–80.
[11] C. Bird, T. Menzies, and T. Zimmermann, “The art and science of analyzing software data,” The Art and Science of Analyzing
Software Data, pp. 1–648, 2015, doi: 10.1016/C2012-0-07289-4.
[12] R. Avros et al., “Boosted decision trees for behaviour mining of concurrent programmes,” Concurrency and Computation:
Practice and Experience, vol. 29, no. 21, Aug. 2017, doi: 10.1002/cpe.4268.
[13] K. Wang, L. Liu, C. Yuan, and Z. Wang, “Software defect prediction model based on LASSO–SVM,” Neural Computing and
Applications, vol. 33, no. 14, pp. 8249–8259, May 2021, doi: 10.1007/s00521-020-04960-1.
[14] R. Malhotra and M. Khanna, “Threats to validity in search-based predictive modelling for software engineering,” IET Software,
vol. 12, no. 4, pp. 293–305, Aug. 2018, doi: 10.1049/iet-sen.2018.5143.
[15] C. Manjula and L. Florence, “Deep neural network based hybrid approach for software defect prediction using software metrics,”
Cluster Computing, vol. 22, no. S4, pp. 9847–9863, Jan. 2019, doi: 10.1007/s10586-018-1696-z.
[16] T. H. Chen, W. Shang, M. Nagappan, A. E. Hassan, and S. W. Thomas, “Topic-based software defect explanation,” Journal of
Systems and Software, vol. 129, pp. 79–106, Jul. 2017, doi: 10.1016/j.jss.2016.05.015.
[17] M. Jorayeva, A. Akbulut, C. Catal, and A. Mishra, “Machine learning-based software defect prediction for mobile applications: a
systematic literature review,” Sensors, vol. 22, no. 7, Mar. 2022, doi: 10.3390/s22072551.
[18] A. Abdu, Z. Zhai, R. Algabri, H. A. Abdo, K. Hamad, and M. A. Al-antari, “Deep learning-based software defect prediction via
semantic key features of source code—systematic survey,” Mathematics, vol. 10, no. 17, Aug. 2022, doi: 10.3390/math10173120.
[19] E. N. Akimova et al., “A survey on software defect prediction using deep learning,” Mathematics, vol. 9, no. 11, May 2021, doi:
10.3390/math9111180.
[20] G. Giray, K. E. Bennin, Ö. Köksal, Ö. Babur, and B. Tekinerdogan, “On the use of deep learning in software defect prediction,”
Journal of Systems and Software, vol. 195, Jan. 2023, doi: 10.1016/j.jss.2022.111537.
[21] S. R. Aziz, T. A. Khan, and A. Nadeem, “Efficacy of inheritance aspect in software fault prediction—A survey paper,” IEEE
Access, vol. 8, pp. 170548–170567, 2020, doi: 10.1109/ACCESS.2020.3022087.
[22] H. Cao, “A systematic study for learning-based software defect prediction,” Journal of Physics: Conference Series, vol. 1487,
no. 1, Mar. 2020, doi: 10.1088/1742-6596/1487/1/012017.
[23] M. H. Mahmud, M. T. H. Nayan, D. M. N. A. Ashir, and M. A. Kabir, “Software risk prediction: systematic literature review on
machine learning techniques,” Applied Sciences, vol. 12, no. 22, Nov. 2022, doi: 10.3390/app122211694.
[24] L. H. Son, N. Pritam, M. Khari, R. Kumar, P. T. M. Phuong, and P. H. Thong, “Empirical study of software defect prediction: a
systematic mapping,” Symmetry, vol. 11, no. 2, Feb. 2019, doi: 10.3390/sym11020212.
[25] A. Kotte and D. A. Moiz Qyser, “A survey of different machine learning models for software defect testing,” European Journal of
Molecular & Clinical Medicine, vol. 7, no. 9, 2021.
[26] M. A. Khan et al., “Software defect prediction using artificial neural networks: a systematic literature review,” Scientific
Programming, vol. 2022, pp. 1–10, May 2022, doi: 10.1155/2022/2117339.
[27] S. Pal and A. Sillitti, “Cross-project defect prediction: A literature review,” IEEE Access, vol. 10, pp. 118697–118717, 2022,
doi: 10.1109/ACCESS.2022.3221184.

 ISSN: 2088-8708
1926
[28] S. Aftab et al., “A cloud-based software defect prediction system using data and decision-level machine learning fusion,”
Mathematics, vol. 11, no. 3, Jan. 2023, doi: 10.3390/math11030632.
[29] A. Alazba and H. Aljamaan, “Software defect prediction using stacking generalization of optimized tree-based ensembles,”
Applied Sciences, vol. 12, no. 9, Apr. 2022, doi: 10.3390/app12094577.
[30] H. Alsawalqah et al., “Software defect prediction using heterogeneous ensemble classification based on segmented patterns,”
Applied Sciences, vol. 10, no. 5, Mar. 2020, doi: 10.3390/app10051745.
[31] S. R. Aziz, T. Khan, and A. Nadeem, “Experimental validation of inheritance metrics’ impact on software fault prediction,” IEEE
[32] P. R. Bal and S. Kumar, “WR-ELM: weighted regularization extreme learning machine for imbalance learning in software fault
prediction,” IEEE Transactions on Reliability, vol. 69, no. 4, pp. 1355–1375, Dec. 2020, doi: 10.1109/TR.2020.2996261.
[33] T. T. Khuat and M. H. Le, “Ensemble learning for software fault prediction problem with imbalanced data,” International Journal of
Electrical and Computer Engineering (IJECE), vol. 9, no. 4, pp. 3241–3246, Aug. 2019, doi: 10.11591/ijece.v9i4.pp3241-3246.
[34] O. Nyarko-Boateng, A. F. Adekoya, and B. A. Weyori, “Using machine learning techniques to predict the cost of repairing hard
failures in underground fiber optics networks,” Journal of Big Data, vol. 7, no. 1, Aug. 2020, doi: 10.1186/s40537-020-00343-4.
[35] E. Khoudry, A. Belfqih, T. Ouaderhman, J. Boukherouaa, and F. Elmariami, “Real-time fault diagnosis system for high-speed
power system protection based on machine learning algorithms,” International Journal of Electrical and Computer Engineering
(IJECE), vol. 10, no. 6, pp. 6122–6138, Dec. 2020, doi: 10.11591/IJECE.V10I6.PP6122-6138.
[36] D. G. Lee and Y. S. Seo, “Improving bug report triage performance using artificial intelligence based document generation
model,” Human-centric Computing and Information Sciences, vol. 10, no. 1, Jun. 2020, doi: 10.1186/s13673-020-00229-7.
[37] A. Marjuni, T. B. Adji, and R. Ferdiana, “Unsupervised software defect prediction using median absolute deviation threshold
based spectral classifier on signed Laplacian matrix,” Journal of Big Data, vol. 6, no. 1, Sep. 2019, doi: 10.1186/s40537-019-
0250-z.
[38] S. Nasir, M. Croock, and S. Al-Qaraawi, “Software engineering based fault tolerance model for information system in plants
shopping center,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 6, pp. 6664–6672, Dec.
2020, doi: 10.11591/IJECE.V10I6.PP6664-6672.
[39] S. Voutsinas, D. Karolidis, I. Voyiatzis, and M. Samarakou, “Development of a machine-learning-based method for early fault
detection in photovoltaic systems,” Journal of Engineering and Applied Science, vol. 70, no. 1, Apr. 2023, doi: 10.1186/s44147-
023-00200-0.
[40] J. Deng, L. Lu, and S. Qiu, “Software defect prediction via LSTM,” IET Software, vol. 14, no. 4, pp. 443–450, Aug. 2020, doi:
10.1049/iet-sen.2019.0149.
[41] H. S. Munir, S. Ren, M. Mustafa, C. N. Siddique, and S. Qayyum, “Attention based GRU-LSTM for software defect prediction,”
PLoS ONE, vol. 16, no. 3, Mar. 2021, doi: 10.1371/journal.pone.0247444.
[42] T. Hai, J. Zhou, N. Li, S. K. Jain, S. Agrawal, and I. Ben Dhaou, “Cloud-based bug tracking software defects analysis using deep
learning,” Journal of Cloud Computing, vol. 11, no. 1, Aug. 2022, doi: 10.1186/s13677-022-00311-8.
[43] M. Jorayeva, A. Akbulut, C. Catal, and A. Mishra, “Deep learning-based defect prediction for mobile applications,” Sensors,
vol. 22, no. 13, Jun. 2022, doi: 10.3390/s22134734.
[44] A. B. Farid, E. M. Fathy, A. S. Eldin, and L. A. Abd-Elmegid, “Software defect prediction using hybrid model (CBIL) of
convolutional neural network (CNN) and bidirectional long short-term memory (Bi-LSTM),” PeerJ Computer Science, vol. 7,
pp. 1–22, Nov. 2021, doi: 10.7717/peerj-cs.739.
[45] Z. Luo, W. Wang, and C. Cen, “Improving bug localization with effective contrastive learning representation,” IEEE Access,
vol. 11, pp. 32523–32533, 2023, doi: 10.1109/ACCESS.2022.3228802.
[46] I. Maduako et al., “Deep learning for component fault detection in electricity transmission lines,” Journal of Big Data, vol. 9,
no. 1, Jun. 2022, doi: 10.1186/s40537-022-00630-2.
[47] C. Pan, M. Lu, B. Xu, and H. Gao, “An improved CNN model for within-project software defect prediction,” Applied Sciences,
vol. 9, no. 10, May 2019, doi: 10.3390/app9102138.
[48] O. Al Qasem, M. Akour, and M. Alenezi, “The influence of deep learning algorithms factors in software fault prediction,” IEEE
[49] H. Wang, W. Zhuang, and X. Zhang, “Software defect prediction based on gated hierarchical LSTMs,” IEEE Transactions on
Reliability, vol. 70, no. 2, pp. 711–727, Jun. 2021, doi: 10.1109/TR.2020.3047396.
[50] W. Yao, M. Shafiq, X. Lin, and X. Yu, “A software defect prediction method based on program semantic feature mining,”
Electronics, vol. 12, no. 7, Mar. 2023, doi: 10.3390/electronics12071546.
[51] T. N. Tengku Asmawi, A. Ismail, and J. Shen, “Cloud failure prediction based on traditional machine learning and deep learning,”
Journal of Cloud Computing, vol. 11, no. 1, Sep. 2022, doi: 10.1186/s13677-022-00327-0.
[52] E. Borandag, “Software fault prediction using an RNN-based deep learning approach and ensemble machine learning techniques,”
Applied Sciences, vol. 13, no. 3, Jan. 2023, doi: 10.3390/app13031639.
[53] A. Khalid, G. Badshah, N. Ayub, M. Shiraz, and M. Ghouse, “Software defect prediction analysis using machine learning
techniques,” Sustainability, vol. 15, no. 6, Mar. 2023, doi: 10.3390/su15065517.
[54] W. Zhang, G. Han, J. Wang, and Y. Liu, “A BP neural network prediction model based on dynamic cuckoo search optimization
algorithm for industrial equipment fault prediction,” IEEE Access, vol. 7, pp. 11736–11746, 2019, doi:
10.1109/ACCESS.2019.2892729.
[55] Y. Z. Bala, P. Abdul Samat, K. Y. Sharif, and N. Manshor, “Improving cross-project software defect prediction method through
transformation and feature selection approach,” IEEE Access, vol. 11, pp. 2318–2326, 2023, doi:
10.1109/ACCESS.2022.3231456.
[56] Y. Hassouneh, H. Turabieh, T. Thaher, I. Tumar, H. Chantar, and J. Too, “Boosted whale optimization algorithm with natural
selection operators for software fault prediction,” IEEE Access, vol. 9, pp. 14239–14258, 2021, doi:
10.1109/ACCESS.2021.3052149.
[57] S. Y. Lee, W. E. Wong, Y. Li, and W. C. C. Chu, “Software fault-proneness analysis based on composite developer-module
networks,” IEEE Access, vol. 9, pp. 155314–155334, 2021, doi: 10.1109/ACCESS.2021.3128438.
[58] Y. Li, W. Eric Wong, S. Y. Lee, and F. Wotawa, “Using tri-relation networks for effective software fault-proneness prediction,”
IEEE Access, vol. 7, pp. 63066–63080, 2019, doi: 10.1109/ACCESS.2019.2916615.
[59] K. Phung, E. Ogunshile, and M. Aydin, “Error-type - a novel set of software metrics for software fault prediction,” IEEE Access,
vol. 11, pp. 30562–30574, 2023, doi: 10.1109/ACCESS.2023.3262411.
[60] I. Tumar, Y. Hassouneh, H. Turabieh, and T. Thaher, “Enhanced binary moth flame optimization as a feature selection algorithm
to predict software fault prediction,” IEEE Access, vol. 8, pp. 8041–8055, 2020, doi: 10.1109/ACCESS.2020.2964321.

1927
[61] J. Zhang et al., “CDS: a cross-version software defect prediction model with data selection,” IEEE Access, vol. 8, pp. 110059–
110072, 2020, doi: 10.1109/ACCESS.2020.3001440.
[62] D. Bowes, T. Hall, and J. Petrić, “Software defect prediction: do different classifiers find the same defects?,” Software Quality
Journal, vol. 26, no. 2, pp. 525–552, Feb. 2018, doi: 10.1007/s11219-016-9353-3.
[63] S. K. Pandey, D. Rathee, and A. K. Tripathi, “Software defect prediction using K-PCA and various kernel-based extreme learning
machine: an empirical study,” IET Software, vol. 14, no. 7, pp. 768–782, Dec. 2020, doi: 10.1049/iet-sen.2020.0119.
[64] Y. Chen and H. Dai, “Improving cross-project defect prediction with weighted software modules via transfer learning,” Journal of
Physics: Conference Series, vol. 2025, no. 1, Sep. 2021, doi: 10.1088/1742-6596/2025/1/012100.
[65] M. Mustaqeem and M. Saqib, “Principal component based support vector machine (PC-SVM): a hybrid technique for software
defect detection,” Cluster Computing, vol. 24, no. 3, pp. 2581–2595, Apr. 2021, doi: 10.1007/s10586-021-03282-8.
[66] B. Marapelli, A. Carie, and S. M. N. Islam, “Software defect prediction using ROS-KPCA stacked generalization model,” in
Smart Innovation, Systems and Technologies, vol. 326, Springer Nature Singapore, 2023, pp. 587–597.
[67] H. K. Dam et al., “Lessons learned from using a deep tree-based model for software defect prediction in practice,” in IEEE
International Working Conference on Mining Software Repositories, May 2019, pp. 46–57, doi: 10.1109/MSR.2019.00017.
[68] S. Hosseini, B. Turhan, and M. Mäntylä, “A benchmark study on the effectiveness of search-based data selection and feature
selection for cross project defect prediction,” Information and Software Technology, vol. 95, pp. 296–312, Mar. 2018, doi:
10.1016/j.infsof.2017.06.004.
[69] L. Sikic, A. S. Kurdija, K. Vladimir, and M. Silic, “Graph neural network for source code defect prediction,” IEEE Access,
vol. 10, pp. 10402–10415, 2022, doi: 10.1109/ACCESS.2022.3144598.
[70] J. Li, P. He, J. Zhu, and M. R. Lyu, “Software defect prediction via convolutional neural network,” in 2017 IEEE International
Conference on Software Quality, Reliability and Security (QRS), Jul. 2017, pp. 318–328, doi: 10.1109/QRS.2017.42.
[71] D. L. Miholca, G. Czibula, and I. G. Czibula, “A novel approach for software defect prediction through hybridizing gradual
relational association rules with artificial neural networks,” Information Sciences, vol. 441, pp. 152–170, May 2018, doi:
10.1016/j.ins.2018.02.027.
[72] A. Iqbal et al., “Performance analysis of machine learning techniques on software defect prediction using NASA datasets,”
International Journal of Advanced Computer Science and Applications, vol. 10, no. 5, pp. 300–308, 2019, doi:
10.14569/ijacsa.2019.0100538.
[73] J. Pachouly, S. Ahirrao, and K. Kotecha, “SDPTool: a tool for creating datasets and software defect predictions,” SoftwareX,
vol. 18, Jun. 2022, doi: 10.1016/j.softx.2022.101036.
[74] Z. M. Zain, S. Sakri, and N. H. A. Ismail, “Application of deep learning in software defect prediction: systematic literature review
and meta-analysis,” Information and Software Technology, vol. 158, Jun. 2023, doi: 10.1016/j.infsof.2023.107175.
[75] P. R. Bal, “Cross project software defect prediction using extreme learning machine: an ensemble based study,” in Proceedings of
the 13th International Conference on Software Technologies, 2018, pp. 354–361, doi: 10.5220/0006886503540361.
[76] A. Bahaa, E. M. Fathy, A. S. Eldin, and L. A. Abd-Elmegid, “A systematic literature review of software defect prediction using
deep learning,” Journal of Computer Science, vol. 17, no. 5, pp. 490–510, May 2021, doi: 10.3844/jcssp.2021.490.510.
[77] I. Batool and T. A. Khan, “Software fault prediction using deep learning techniques,” Software Quality Journal, vol. 31, no. 4,
pp. 1241–1280, Jun. 2023, doi: 10.1007/s11219-023-09642-4.
[78] C. Pan, M. Lu, and B. Xu, “An empirical study on software defect prediction using CodeBERT model,” Applied Sciences, vol. 11,
no. 11, May 2021, doi: 10.3390/app11114793.
[79] R. Malhotra and S. Kamal, “An empirical study to investigate oversampling methods for improving software defect prediction
using imbalanced data,” Neurocomputing, vol. 343, pp. 120–140, May 2019, doi: 10.1016/j.neucom.2018.04.090.
[80] A. Alsaeedi and M. Z. Khan, “Software defect prediction using supervised machine learning and ensemble techniques: a
comparative study,” Journal of Software Engineering and Applications, vol. 12, no. 5, pp. 85–100, 2019, doi:
10.4236/jsea.2019.125007.
BIOGRAPHIES OF AUTHORS
Deepti Rai completed her master degree in 2016 and bachelor degree in 2004
from Visvesvaraya Technological University, India. She is currently pursuing her doctoral
degree in the domain of machine learning at the Department of Computer Science and
Engineering, Ramaiah University of Applied Sciences, Ramaiah Technology Campus,
Bengaluru, Karnataka, India. She has 10 years of experience in teaching and 6 years of
Industry experience. Her research interest is in the field of machine learning, deep learning,
AI, and cloud computing. She can be contacted at email: deeraisecond@gmail.com.
Jyothi Arcot Prashant completed her Ph.D. in 2020, master degree in 2009,
Bachelor degree in 2002 from Visvesvaraya Technological University, India. She is currently
working as Faculty of Engineering and Technology, Department of Computer Science and
Engineering, Ramaiah University of Applied Sciences, Ramaiah Technology Campus,
Bengaluru, Karnataka, India. She has 16 years of experience in teaching and has published
many research papers in journals indexed in SCI/SCIE, WOS, SCOPUS, and presented papers
in several National and International conferences. Her research interest is in the field of
wireless sensor network, IoT, embedded systems, AI, ML and deep learning. She can be
contacted at email: jyothiarcotprashant@gmail.com.

Insights of effectivity analysis of learning-based approaches towards software defect prediction

More Related Content

Similar to Insights of effectivity analysis of learning-based approaches towards software defect prediction (20)

More from IJECEIAES (20)

Recently uploaded (20)

Insights of effectivity analysis of learning-based approaches towards software defect prediction