Discover Computing

Review

A comprehensive survey on intelligent software reliability prediction

Ajit Kumar Behera1 · Pamela Chaudhury1 · Ch. Sanjeev Kumar Dash1

Received: 20 September 2024 / Accepted: 7 May 2025

© The Author(s) 2025  OPEN

Abstract
The modern world is propelled by the evolution of software systems. The application of software systems is currently
ubiquitous in areas such as process control, business, transportation, medical, housing and building automation, and
government applications. Reliable software consistently performs its intended functions without failure, producing
accurate and expected results under predefined conditions. Reliability in software is crucial because it impacts user
trust, system performance, and overall software quality. In this context, we extensively surveyed 140 research papers
from 2005 to 2024 in the domain of intelligent systems for software reliability prediction. This survey aimed to study and
analyze the robust techniques for reliable software development used by different researchers. Additionally, we explored
the advantages and limitations associated with software reliability prediction methods. We have also highlighted the
challenges involved in SRP and examined suggested strategies aimed at enhancing these predictions. After a compara-
tive analysis, we found that the hybrid methods provided the best results for software reliability prediction. This in turn
would help to gain a competitive advantage in the market, in developing high-quality software.

Keywords Software reliability · Software quality · SRGM · Neural networks · Fuzzy logic · Genetic programming ·
Evolutionary algorithm · Deep learning

1 Introduction

Software systems are essential to modern-day societal routines. The applicability of software has now spread to a wide
range of areas, including home appliances, social networking, banking, automotive, aerospace, medical, nuclear, etc.
To enhance productivity and performance, most commercial and government organizations rely on software systems.
With the increase in software size, complexity, and demand, more reliance on software has increased. As a result, the
likelihood of failures occurring due to software faults has increased.
The Standish Group has been tracking IT project outcomes and compiling a database of 50,000 projects since 1994.
According to the Standish Group Chaos research, projects are classified as successful (finished on time and within budget),
challenging (not completed on time and within budget), or failed (canceled or not completed) [1]. From the report, it
is observed in 2020 that, only 35% of total projects were fully successful, whereas 19% of the software projects were
canceled before completion and failed. Most software projects fail due to poor software quality. Failure of the software
system leads to user dissatisfaction, inconvenience to social life, severe revenue loss, and, loss of life. Some of the exam-
ples of recent software failures that have resulted in devastating outcomes are [2]: (i) The Iowa Caucus Smartphone app:
In 2020, a smartphone app was created to aid in the announcement of the Iowa caucus results in the United States.

* Ajit Kumar Behera, [email protected]; Pamela Chaudhury, [email protected]; Ch. Sanjeev Kumar Dash, sanjeev_dash@yahoo.
com | 1Department of Computer Science & Engineering, Silicon University, Bhubaneswar 751024, Odisha, India.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

However, the app had not been well tested in its intended usage setting, and its installation was not straightforward.
This was certainly a system and process failure, with coding issues. The damage done to the Iowa caucus’credibility was
immeasurable. (ii) British Airways glitch: Over one hundred British Airways flights were canceled on August 7, 2019,
while another 300 were delayed. The application malfunctioned during the busiest month for air travel. Thousands of
travelers were forced to stay behind and wait in overcrowded airports for long periods. (iii) WannaCry: Due to software
faults, a significant ransomware assault known as WannaCry has shattered the trust model on which ransomware relies
to be paid, potentially putting an end to ransomware as one of the most common malware methods. (iv) Welsh NHS IT
failure: Due to a software failure in 2018, doctors and hospital employees in the Wales NHS were incapable of accessing
patient information such as blood and X-ray findings. Patients were unable to cancel appointments, and notes could
not be written or stored in the NHS system, leading to a backlog. v) O2: More than 30 million O2 customers in the UK lost
data access on December 6 th, 2018, due to a software malfunction that prevented them from using 3G and 4G services.
(vi) Therac-25: A radiation therapy machine that was operated by a computer. Six patients died because of software fail-
ures owing to therapeutic overdose. As a result, the industry and the users of different software applications worldwide
demand not only technically accurate software but also safe and reliable software.
The reliability of a software product fundamentally signifies credibility or dependability. It is defined as “The prob-
ability of failure-free software execution for a specified period in a specified environment” [3]. The failure probability F(t)
is correlated to reliability R(t), which is described in Equation 1.
R(t) = 1 − F(t) (1)
Different terms such as error, fault, and failure are associated with software reliability problems [4]. An error is an
associate degree action created by someone that leads to an inaccurate result. A fault is the manifestation of a program-
ming error. A fault is a defect in a program that causes it to fail when it is run under specific conditions. Identifying and
resolving the defects that cause failures have a significant impact on software reliability. In the software industry, qual-
ity assurance of software depends on assessing software reliability [5]. Software reliability estimation analyses failure
data gathered during testing. Software reliability is difficult to assess due to software complexity. As a result, various
techniques for predicting software failures over time have been developed. Regardless of users’ perspectives, software
bugs are the apparent source of reliability issues. Evaluating software reliability comprises two distinct tasks: estimation
and prediction [6–8]. Recently, van Driel et al. [9] introduced a predictive model tailored to estimate software reliability
within agile testing settings. Recently, a new mathematical model was proposed by Diwaker et al. [10] for the estima-
tion of the reliability of software using series and parallel reliability approaches and it was found that the CBSE reliability
model has a lower error rate.
By leveraging intelligent systems to predict software reliability, development teams can make data-driven decisions
that improve product quality while saving time and resources. This approach ultimately supports faster, more efficient
development cycles, helping teams deliver reliable software with fewer hours spent on reactive debugging and testing.
Researchers have used intelligent systems to handle nonlinear, time-varying, and complex real-world problems with great
success. In the present context, intelligent systems can offer a positive viewpoint on software reliability prediction. To
design software reliability models, researchers have recently used intelligent systems including Fuzzy Logic (FL), Artificial
Neural Networks (ANN), and Genetic Programming (GP) [11, 12].
The primary aim of this paper is to highlight the development of different software reliability prediction models
over time, leading up to the present day. In the literature there exists several State-Of-The-Art (SOTA) survey papers
in the field of software reliability prediction. Generally, SOTA survey paper provides a comprehensive overview of the
latest advancements in the field of software reliability prediction, which covers both parametric and non-parametric
methods. It covers broader categories of methods including statistical models, regression analysis, and expert sys-
tems. They present comparative analysis of different approaches in terms of accuracy, efficiency, and applicability.
In our survey paper, we focused on an intelligent system-based software reliability prediction model. It provides an
in-depth review of various ML algorithms such as neural networks, swarm intelligence, fuzzy logic, deep learning,
ensemble, and hybrid techniques. In this paper, we review and examine the literature on software reliability concern-
ing existing software reliability prediction models and available metrics. It offers a comprehensive and sophisticated
examination of the progression of computational machine learning models used in predicting software reliability. The
research gaps in the recent software failure mitigation techniques have been identified. In enhancing the study, an
exploration of both the advantages and limitations of each model is conducted as well. Ultimately, this survey aims

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

to steer researchers toward constructing mitigation models for both traditional and advanced software reliability
prediction, offering valuable guidance in the process.
This survey follows a systematic approach to reviewing existing literature on software reliability prediction.
Research articles were collected from reputable digital libraries, including IEEE Xplore, ACM Digital Library, Springer,
and Elsevier, using keywords such as “software reliability prediction”, “machine learning in software reliability,” and
“failure prediction models”. Only peer-reviewed papers published in the last two decades were considered to ensure
relevance. The selected studies were categorized based on prediction techniques, including machine learning
approaches, and hybrid techniques. Furthermore, this survey evaluates key performance metrics such as average
error (AE), magnitude relative error (MRE), mean absolute error (MAE), mean squared error (MSE), and normalized root
mean squared error (NRMSE) to compare various methodologies. The findings were synthesized through systematic
analysis and tabular comparisons to highlight trends, challenges, and future research directions in software reliability
prediction. In our selection funnel, we used the following related keywords: software reliability, software defects,
swarm intelligence, genetic algorithm, fuzzy logic, swarm algorithm, deep learning, hybrid model, and survey. These
keywords were combined using Boolean expressions"AND"and"OR."All studies cited and referenced in the survey
were sourced from academic journals or conferences.
Section 1 describes the introductory concepts of software reliability prediction and its motivation. Section 2
briefly categorizes software reliability models. Section 3 describes the various intelligent system-based software
reliability prediction methods in detail. In Sect. 4, the survey conducts a critical analysis and investigation, while
Sect. 5 concludes this review work.

2 Software reliability models

A software reliability model (SRM) assesses software reliability and provides crucial insights for project managers.
It includes details such as the number of remaining errors after a specified testing period and the time required to
reach a predefined reliability target [13]. Additionally, these models emphasize discovering the key software reliabil-
ity challenges and the factors that impact them. This will enhance the cost-effectiveness and timeliness of software
development. Since reliability is regarded as the"must-have quality"aspect of a product, several SRMs have been
proposed and implemented. A survey on the prediction of software reliability was conducted by Rastogi et al. [14]
using various soft computing techniques. Recently, Oveisi et al. [15] provided insight into the categorization of diverse
techniques for forecasting software reliability. To increase the accuracy of such forecasts, they have also discussed
the difficulties associated with prediction methods and offered some partial answers (i.e., Bayesian methods). Kong
et al. [16] have conducted a systematic review in the field of software reliability prediction using swarm intelligence
algorithms. A multitude of software reliability models were found in the literature, with new ones continually emerg-
ing each year. Classification of different software reliability models is illustrated in Figure 1.

Fig. 1  Types of software reli-

ability model

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

2.1 Software reliability growth models

A software product undergoes a series of identifiable stages during its lifetime. During the testing phase of a software
process, different errors causing software failures are detected and corrected. As many errors are corrected, the reliability
of the software gradually improves. This process of increasing reliability is regarded as reliability growth. An SRGM pos-
tulates that software reliability continuously increases as the detected errors are corrected during software testing [17,
18]. The reliability growth models, which are black-box models in nature, have been used to find and remove defects in
software. These models consider the failure frequency and cumulative number of failures. To assess reliability, cumulative
failures or the time between failures that persist after testing are considered in these models.
The SRGMs are divided into concave and S-shaped models. Figure 2 shows both types of models. In concave mod-
els, the underlying assumption is that the error/defect detection is directly related to the defects present in the code.
Consequently, as more defects are identified and subsequently repaired, the defect detection rate gradually diminishes.
Conversely, S-shaped models suggest that in the initial testing phases, fewer defects are identified because of new team
members. As testing progresses, the team’s familiarity with the environment leads to a higher defect detection rate in
the later stages. In the scenario of imperfect debugging, the software’s fault content remains the same. However, during
error generation the fault content increases as testing advances. Kapur et al. [19] introduced two general frameworks for
developing various software reliability growth models using a non-homogeneous Poisson process (NHPP). It is accounted
for both imperfect debugging and error generation.

2.2 Early software reliability models

Software development involves various identifiable stages, popularly known as the software development life cycle
(SDLC). Typically, faults are induced in almost every phase of the software process and continuously propagate to the
next stage until otherwise identified. Predicting the number of faults in the requirements, design, and coding process
might lead to mitigation activities such as more reviews and more rigorous testing. Generally, software failure informa-
tion is not available at the early stages, and numerous metrics related to reliability are utilized in the preliminary stages
of SDLC. For both software developers and managers, the early software reliability model is important because it helps in
the early assessment of software quality, cost overruns, and errors, as well as the most effective development techniques
[20]. IEEE published a standard guide in 1998 called “IEEE guide for the use of IEEE standard dictionary of measures in
the development of reliable software”. This standard is used by project managers, and end-users to ensure that software
products are as reliable as possible.
Early, in the software design phase, the fuzzy logic toolkit was used to provide a method for measuring the error
susceptibility and density of modules. Rathi et al. [21] proposed an effective Software Fault Prediction (SFP) model by
efficiently addressing class imbalance problems. Their proposed models were evaluated using eight different classifiers,
and their performance was evaluated on fifty-four open-source datasets.
Accurate prediction of software defects in different stages of the software process is crucial for the software profes-
sionals to determine the release time of software [22–24]. Pandey and Goyal [25] employed fuzzy set theory and the CMM
level to capture the faults in the initial phases of software development. Samal and Kumar [26] used expert opinion and
employed an extended ANN model for early software reliability prediction by reusing past data from related projects.
For the software to perform properly and to reduce the resources and time needed for future maintenance, it is crucial
to find software bugs as soon as feasible. Particle swarm optimization (PSO) and the random forest (RF) classifier were
used in a method by Santosh et al. [27] to increase the defect prediction rate in software.

Fig. 2  S-shaped and concave

model

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

Early software failure prediction is a vital activity that might result in major problems including underestimation and
overestimation. Rajput et al. [28] developed two alternative methods for fault prediction at an early stage utilizing an
analogy-based approach. First, they created a complete hybrid model that combined regression and genetic algorithms.
In the second approach, the best-fit reduced parameters for handling uncertainty were combined with classification
data from neural networks to create clustering. Rao and Reddy [29] employed an under-sampling strategy to classify
imbalanced software datasets, aiming to improve the accuracy in predicting error-prone modules with reduced time
and simplicity.

2.3 Architecture‑based software reliability models

Software reliability models utilizing software faults/defects are referred to as black-box approaches. The key characteristic
of these black-box models is the use of stochastic methods to represent the failure process, for the cumulative number
of failures over a specified time [30]. Several architecture-based software reliability models were also discussed in [31,
32]. The difficulty with these models that these systems just interacted with the outside world and ignored its internal
design. Typically, these models were used to analyze or forecast software reliability that does not use any knowledge
apart from software failure data. As a result, black-box models are unsuitable for modeling a huge component-based
software system with several software packages. As a result, software reliability must be evaluated using a white-box
technique that considers the internal design of the product. Architecture-based models can be used during the design
and implementation phases of software development. These models can aid in the identification of important compo-
nents and interfaces that can aid in more effective design decisions.

2.4 Bayesian models

The Bayesian model emphasizes that observing software behavior is more crucial than estimating the number of faults
present. The model suggests that if no failures are observed, the software’s reliability should improve, reflecting the user’s
increased confidence. Thus, reliability is viewed as a function of both the number of detected faults and the duration
of fault-free operation [33]. Generally, the reliability assessment of a new system depends on the failure history of pre-
existing systems. Littlewood et al. [34] introduced a Bayesian method for assessing software reliability by combining the
reliability data from an older system. To predict reliability, several Bayesian models have been proposed that use prior
information about the system as well as available software failure data [35–37].

3 Software reliability prediction techniques

Software reliability models are used for measuring failure intensity, cumulative failures, failure time intervals, faulty
modules, and cost estimation [38, 39]. Software experts can assess the reliability of software by calculating cumulative
failures and successive failure times. Several software modules are integrated to create large and sophisticated software
systems. None of the modules are similarly significant or have the same number of faults. Therefore, researchers began
to differentiate between fault-prone (FP) and non-fault-prone (NFP) software modules. Software quality metrics may
be used to identify a module’s fault-proneness. Some measurements are connected and redundant. To forecast defect
pone modules, it is necessary to determine relevant software metrics. Project managers often seek information about
the cumulative errors present in a product after a certain period of testing to determine the software release schedule.
According to modeling theory, software reliability models can be classified into a) parametric models and b) non-
parametric models. Parametric models are associated with a set of assumptions about how software is developed and
tested, stochastic behavior based on the testing process, and the nature of software products [40]. The complexity of
a software system increases exponentially with its size. The parametric model faces problems such as low adaptivity to
complicated, uncertain, and changing environments of software [41].
To address the limitations of parametric models, the researchers often employ non-parametric models [42]. Non-
parametric models, appear to offer greater predictive performance than parametric models. Nonparametric models
are based on intelligent systems which include support vector machine (SVM), artificial neural network (ANN), fuzzy
logic (FL), and genetic algorithm (GA) [43–45]. These models do not make such assumptions, instead use a time-series
approach such as moving average (MA), auto-regressive (AR), ANN, deep learning, Markov chain, genetic algorithms,
and fuzzy logic [46, 47]. Liapis et al. [48] carried out a comparative study between active learning-enhanced models and

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

voting ensembles to improve certain aspects of classification tasks in software defect prediction. Recently, many other
researchers conducted a comprehensive review of software reliability prediction techniques using intelligent systems
[49–51]. In this subsection, we have discussed different intelligent system techniques applied for software reliability
prediction in Figure 3.

3.1 ANN‑based software reliability prediction

Artificial neural networks (ANNs) are comprised of neurons, which are fundamental processing units capable of storing
and applying knowledge gained through experience. ANNs emulate the brain’s information-processing capabilities.
ANN can be applied across diverse domains like image recognition, natural language processing, speech recognition,
and financial forecasting. They can generalize well to unseen data, making them robust in predictive tasks. Additionally,
they offer superior predictive accuracy for software reliability assessment. The capacity of ANNs to approximate univer-
sal functions has prompted researchers to use them in SRP [52]. The elementary structure of the ANN-based prediction
model is depicted in Fig. 4. The time series approach is employed to predict the intervals between software failures. A
vector of previous time between failure data (x1 , x2 , x3 … , xp ) serves as the input, while xp+1 represents the next failure
time as the output of the artificial neural network (ANN).
ANN has gained considerable importance in predicting software reliability since it requires only software failure data
from the system [53]. These models make no assumptions about software development methodologies, software test-
ing processes, or software project features. ANNs can roughly estimate any non-linear continuous function. As a result,
ANN-based software reliability prediction models are becoming increasingly popular. Karunanithi et al. [54] were the
first to utilize an ANN model for predicting software reliability. Soon after, RBF neural networks (RBFNNs), MLP, FLANN,
and regression neural networks were [55, 56] used in software reliability prediction.
To handle complex problems, it is essential to develop high-performance complex software systems. Since differ-
ent software exhibits different behaviors, no single model provides accurate software predictions in all the cases. With
this motivation, Vishwanath [57] proposed a set of neural network models in his doctoral dissertation and used neural
networks for software reliability prediction. Kaur et al. [58] proposed an ANN-based model for improved estimation of
software effort for NASA software projects. Wu and Huang [59] compared neural networks and parametric recalibration
models for the prediction of software reliability and observed that neural networks are substantially greater predictors.
Cai et al. [60] used an MLP (multilayer perceptron) model to handle two software failure datasets and revealed that the
network generates satisfactory results for dynamic software failure data. It was also observed that an MLP has a slower
learning rate than a feed-forward neural network (FFNN). In 1986, Rumelhart et al. [61] established the back-propagation

Fig. 3  Intelligent system-

based SRP techniques

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

Fig. 4  Basic structure of the

ANN-based prediction model

learning method for training neural networks. With this motivation, Sigh and Kumar [62] shown that an FFNN using the
back-propagation algorithm can effectively and significantly predict software reliability growth trends. However, one
of the limitations of this approach is that it has a slower learning rate than an FFNN. During neural network training, the
gradient descent method is typically employed for optimizing the weights and biases. Furthermore, the gradient descent
search process creates problems such as solution trapping in local minima, slow convergence, activation function satura-
tion, weight interference, overfitting, and poor generalization capabilities from noisy datasets. Additionally, it is difficult
to obtain the optimum network design and the ideal number of hidden units [63]. Subsequently, many researchers have
used alternative approaches to handle these issues.
Arya et al. [64] investigated an ML-based K-means clustering technique for software defect prediction. The developed
prediction model integrates the OOPS metric feature-based software project dataset, including fault feature specifica-
tions. The final stage involves clustering the connected metric attributes using K-means clustering. To assess software
quality, an ANN using the back-propagation technique was utilized [65, 66]. Furthermore, researchers have shown inter-
est in semiautomated intelligent estimators including SVMs, ELMs, FLNNs, MLPs, and RBFNs for the SRP [67–69]. Using
artificial neural networks, Lo [70] developed a model for failure detection, and fault rectification for the SRP. This method
looks at numerous traditional software reliability growth models without making any unreasonable assumptions. The
use of several layers may result in a complex ANN structure that leads to overfitting.
Pai and Hong [4] introduced a support vector machine-based software reliability prediction model. To choose param-
eters for their SVM model, they employed the simulated annealing technique. The findings revealed that their predictions
were more accurate than those of existing methods for determining the SRP. Jin [71] introduced an SVM-based software
reliability prediction model. He combined the genetic algorithm with the simulated annealing process to create GA-SA,
a new algorithm.
Using a next-step-predictability model, Tian et al. [72] proposed an evolutionary connectionist technique for fail-
ure time prediction. The model outperformed certain previous ANN models in terms of next-step prediction. Subse-
quently, Bal et al. [73] proposed a nonparametric approach employing an RBFN for software fault prediction. Rath et al.
[74] proposed an extreme learning machine (ELM) approach for software reliability prediction. Furthermore, the article
explored the application of the ELM in the process of feature selection for software reliability prediction. The testing
phase utilized datasets from NASA’s Metrics Data Program. The model presented combines ELM with feature selection,
and demonstrated notably improved prediction accuracy compared to conventional ELM models for defect prediction. To
improve generalization and reduce the risk of overfitting, the proposed model incorporated the Bayesian regularization
method. Recently, Ndahi et al. [75] proposed a novel model known as the Pi-sigma neural network (PSNN) to enhance

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

comprehension of the modeling of defects and reliability within software systems. The model’s effectiveness was verified
using five datasets from NASA’s PROMISE project. Sometimes for enhanced fault-free software performance, opting for
selected features from a dataset is more advisable than utilizing all available features [76]. Toofani and Garg [77] used
random forest, logistic regression, and SVM techniques with four distinct feature selection techniques (correlation coef-
ficient, Fisher score, LASSO, and recursive feature selection). The outcomes demonstrated that the performance of this
model ranged from moderate to satisfactory. Su and Huang [8] utilized neural network-based approaches for software
reliability, developing a dynamic weighted combinational model that yielded improved predictions. Different Software
reliability prediction techniques have been evaluated using different performance metrics such as mean squared error
(MSE), average error (AE), mean relative error (MRE), root mean squared error (RMSE), and normalized root mean squared
error (NRMSE). Table 1 compares the performance of some of the ANN-based software reliability prediction techniques.
From the table, it is observed that different researchers adopted different ANN-based techniques with different datasets.
From the table, it is observed that the performance measures of most of the techniques are based on MSE. Among all
algorithms, the MSE value of the FFNN algorithm is 0.0016 which is the minimum.
Artificial Neural Networks (ANN) are highly effective in modeling complex, non-linear relationships and are capable of
learning and adapting to new data through parallel processing, making them suitable for high-dimensional datasets. They
are also fault-tolerant, handling noisy and incomplete data effectively. Artificial Neural Networks (ANN) are extensively
employed in SRP for their adaptability and ability to learn complex patterns and relationships effectively. However, ANN
models often face challenges such as overfitting, especially when the network is too complex or trained on insufficient
data. Training ANNs can be computationally expensive and time-consuming, requiring significant hardware resources.
Additionally, the"black-box"nature of ANN models makes it difficult to interpret and explain their decision-making pro-
cesses. The common remedy for overfitting problem is to implement various regularization techniques e.g. dropout. For
“black-box” nature of ANN, we can use interpretability tool like LIME or SHAP to understand the model decision.

3.2 Swarm and evolutionary computation (SEC)‑based software reliability prediction

Swarm intelligence (SI) and evolutionary computing (EC) represent two separate lineages of search and global optimiza-
tion techniques within computational intelligence [78].
Such methods can be inspired by biological evolution, fall under the category of EC algorithms, or be the conse-
quence of social insects’collective life-sustaining behaviors, in which case they are referred to as SI algorithms. The
term"Evolutionary Computing"(EC) refers to a group of techniques that use natural selection and Darwinian evolution
to tackle optimization problems [79]. One of the fields of computer science that is expanding the fastest is the use and
development of EC algorithms [80]. The SI denotes the behavior of dispersed, self-organized systems. SI systems are
usually based on a small number of fundamental agents that interact with one another and their environment on a local
level. SI algorithms include genetic programming (GP), particle swarm optimization (PSO), ant colony optimization (ACO),
and gray wolf optimization (GWO).
Genetic programming (GP) is a technology created from GA that is based on biological evolution (GA) [81, 82]. Soft-
ware reliability growth may be modeled using GP. It is a symbolic regression challenge to develop software reliability
growth models using GP [83]. Selecting the appropriate parameters for evolutionary learning algorithms, which is essen-
tial for finding global optima, is a highly challenging task. A novel metaheuristic technique named chemical reaction

Table 1  Performance of ANN-based software reliability prediction techniques

Paper Algorithms Data set MSE AE/RE MRE NRMSE

Lakshmanan and Ramasamy [18] Neural Network Combination model with the Musa data set 0.2953
base models G-O model, Delay S-shape model
Bhuyan et al. [51] BPNN 3.002 0.0272
Jaiswal & Malhotra [52] FFBPNN Project data 231.55 0.0442
Bisi & Goyal [53] Neural Network with Logarithmic Encoding (NLE) Electronic 8.66 1.49
switching
System
Noekhah et al. [56] ICA-MLP DACS data set 0.0516
Singh & Kumar [62] FFNN DACS data set 0.0016 0.0004

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

optimization (CRO) requires fewer learning parameters without compromising predictive capability [84, 85]. This approach
emulates the molecular interactions found in chemical reactions, aiming to attain a state of low energy stability.
A key challenge in an MLP is determining the ideal number of nodes in each layer and the appropriate number of layers
in the network. To overcome these issues, a higher-order neural network (HONN) has evolved [86]. By adding additional
higher-order units to the input layer, represented as functional expansion, the architecture creates an implicitly hidden
layer to account for the nonlinear connection between the input layer and the output layer. FLANNs, RVFLs, and ELMs are
all part of the HONN domain. The functional link neural network (FLANN), known for its flat structure, achieves enhanced
mapping of input-output nonlinear connections by expanding input signals while ensuring reduced computational
complexity. The optimized FLANN models are used to predict the closing prices of Bitcoin on a daily, weekly, and monthly
basis [87]. Motivated by this, Behera et al. [88] used a FLANN optimized by PSO to predict software reliability. According to
their findings, the PSO-FLANN model outperforms conventional models such as the BPNN, FLANN, DENFIS, and NEBPNN.
To design an optimal ANN structure, Behera et al. [89] utilized an artificial electric field algorithm (AEFA) to determine
the best possible weights, biases, and number of hidden neurons. The resulting hybrid model, which combines AEFA
and ANN, is assessed using datasets for modeling and forecasting software reliability. Real software failure datasets are
used in experiments that consider normalized root mean squared error statistics. Sometimes, software failure datasets
feature a diverse distribution of instances, including both majority and minority samples, which affects the predictive
capability of a model. To reduce data dimensionality and remove noisy attributes from software failure datasets, Rahman
et al. [90] employed PSO, focusing on enhancing classification performance in software defect prediction.
The gray wolf optimizer (GWO) stands out among the metaheuristics and scope for its unique ability to emulate the
social hierarchy and hunting behavior of gray wolves. Wang et al. [91] developed the binary gray wolf optimization
(bGWO) algorithm for optimal feature selection and developed a robust defect predictor. Recently, Oueslati and Manita
[92] introduced a novel approach, integrating logistic regression (LR) with the fractional chaotic gray wolf optimizer
(FCGWO), to address the challenges in software defect prediction (SDP). The results demonstrate a significant improve-
ment in defect prediction performance, with the LR-FCGWO model outperforming traditional models in terms of accuracy
and robustness. Malhotra and Khan [93] proposed an SRP model by integrating a modified gray wolf optimization (GWO)
algorithm as a wrapper-based feature selection method with the synthetic minority oversampling technique to balance
the dataset. Their goal was to enhance the prediction efficiency of the learning model.
The common difficulty of ANN-based time series modeling is that large data sets are needed for training a model. Usu-
ally, software failure datasets are very small in size. A technique for expanding training datasets by creating virtual data
points was developed by Nayak et al. [94] for stock market prediction. Motivated by this, Behera et al. [95] predict software
reliability using evolutionary virtual data-position (VDP) investigation. To improve the prediction of software faults, a
multilayer perceptron (MLP)--based virtual data position (DEVDP) technique was developed. The network’s parameters
are optimized using the evolutionary process known as differential evolution (DE). The model was validated in the pres-
ence of eight failure datasets from distinct sources using VDPs. Table 2 compares the performance of SEC-based soft-
ware reliability prediction techniques. From the table, it is observed that most of the works used AE as the performance
measure. It is also observed that the AE value of the GA-based technique is 1.79, which is smaller than other algorithms.
SEC methods perform well on complex, high-dimensional, and non-linear problems where traditional methods strug-
gle. These algorithms are designed to avoid local optima by exploring the search space widely, increasing the likelihood of
finding a global optimum. However, their effectiveness depends on careful design, parameter tuning, and computational
resources. They are particularly suited for problems where traditional optimization methods fail or are impractical. To
address the scalability problem, we can design adaptive algorithm that can scale with problem size. Parameter tuning
can be achieved by integrating some machine learning techniques to optimizes the parameters dynamically.

Table 2  Performance of References Algorithms Data set AE/RE MRE NRMSE

SEC-based software reliability
prediction techniques Tian & Noore [66] Evolutionary Neural Musa data set 1.79
Network (GA)
Roy et al. [79] PSO Musa data set 5.0246 0.0099361
Bisi [80] PSO Musa data set 8.61
Costa et al. [81] GP Musa data set 10.26
Behera et al. [86] CRO + FLANN Musa data set 8.6939e-006
Behera et al. [88] PSO + FLANN Musa data set 0.1251

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

3.3 Fuzzy‑logic‑based software reliability prediction

Fuzzy logic, introduced by Zadeh [96], was developed as a mathematical model to address ambiguity in everyday situ-
ations. Intelligent systems offer alternative approaches to represent and manage uncertain, incomplete, imprecise, or
noisy data. Fuzzy logic is a type of many-valued or probabilistic logic that focuses on approximate reasoning rather than
fixed, exact conclusions. The significance of fuzzy logic emerging from human thinking, particularly common reasoning,
is approximate.
Measuring in software engineering is challenging for two main reasons. First, software engineering emphasizes engi-
neering principles less than mathematics [97]. Second, many software characteristics, such as maintainability, portability,
reliability, defect correction, and handling unusual behavior, are qualitative rather than quantitative. Assessing these
characteristics relies on expert judgment, so it is crucial to enhance these evaluations to achieve consistent and accurate
attributes. The types of scales for attribute measurement are defined as ordinal, nominal, interval, or absolute [98]. How-
ever, choosing the characteristics that reflect the correctness of a software project is a challenging process. Researchers
and software professionals typically investigate the association between software failure times and all variables for which
the data of the researched environment are accessible in the intervals between software failures. Fuzzy logic is suitable
for collecting subjective knowledge about software metrics in the early stages of software development. For all these
reasons, we believe that these are ideal tools for understanding how software is created and how it will behave in real-
world production environments. The developer cannot see most of the early phases of software measurements, and
assigning a crisp to the theme is difficult. As a result, fuzzy logic was found to be an appropriate method for dealing with
these problems. Early fault prediction offers the chance to identify software quality issues, cost overruns, and optimal
development strategies at an early stage. In the early stages of software development, fuzzy logic has proven beneficial
for gathering and interpreting subjective input concerning software metrics. A diagrammatic representation of a fuzzy
logic-based model for software reliability prediction is illustrated in Figure 5. A fuzzy logic system consists of four key
components: the fuzzification process (input), the fuzzy rule base, the fuzzy inference process, and the defuzzification
process (output). Fuzzification involves transforming crisp values into a fuzzy format. Based on the available data and
associated uncertainties, input and output parameters are expressed as linguistic variables such as low (L), medium (M),
and high (H). At the core of any fuzzy system lies the fuzzy rule base. The fuzzy inference process integrates the rules
from the fuzzy rule base to establish a mapping between fuzzy inputs (input memberships) and fuzzy outputs (output
memberships).
In the field of fuzzy logic-based software reliability, many related studies have been published [100–103]. Traditional
software reliability predictions based on single models often fall short in accuracy because software failure patterns
continuously evolve. Shen et al. [103] employed a fuzzy system-based neuro-fuzzy inference system. The SRM aims to
reduce randomness and computational complexity while avoiding local optima and searching for optimal parameters.
Aljahdali et al. [104] proposed a Takagi-Sugeno (TS) fuzzy model for an SRP, where three separate applications were used
for their experimental work. Yadav and Yadav [105] proposed a fuzzy logic-based phase-wise software defect predic-
tion model using the top most reliability-relevant metrics of each phase of the SDLC. To provide clear output, the fuzzy

Fig. 5  Basic structure of the

fuzzy logic-based software
prediction model

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

Table 3  Performance of References Algorithms Data set MSE AE/RE

Fuzzy logic-based software
reliability prediction Chatterjee et al. [101] Fuzzy Logic Musa data set 8.8378
techniques
Roy et al. [102] Fuzzy Logic Musa data set 4.0249
Shen et al. [103] Regularized Fuzzy Real-time command and 0.4 2.8768
learning control

Fig. 6  An ensemble model for

the SRP

reasoning method considers subjective information, where a Bayesian framework is used to generate the crisp output.
Khatatneh [106] suggested a methodology for forecasting software reliability using fuzzy logic and found superior pre-
dictive capability compared with earlier models. Dimov and Punnekkat [107] presented a component-based software
system and reliability prediction using fuzzy logic. For the uncertainty issue, they used possibility theory, which was based
on fuzzy sets. Later, Jaiswal et al. [108] introduced an innovative reliability model using fuzzy logic, employing if-then
rules to assess software reliability based on the code’s component structure. These models can evaluate interdependent
components, operational profiles, reusability, and the overall software reliability of component-based systems. Table 3
compares the performance of fuzzy logic-based software reliability prediction techniques. From the table, it is observed
that the performance measures of most of the algorithms are based on AE. It is also found that the AE value of regular-
ized fuzzy learning techniques is 2.8768 which is smaller than other techniques.
Fuzzy logic is a robust tool for dealing with uncertainty and imprecision, making it invaluable for many real-world
applications. However, its effectiveness depends on the expertise applied during system design and its suitability for
the problem at hand. Sometimes, complex systems can lead to an exponential increase in rules. Also combining fuzzy
systems with other ML models can be challenging. Possible solution for rule explosion problem is to apply rule reduction
techniques to manage the complexity. Integrating fuzzy inference systems with ML we can handle uncertainty effectively.

3.4 Ensembled software reliability prediction techniques

Ensemble models exemplify the power of collective intelligence in machine learning. It has been shown that combining
multiple predictors generally outperforms using a single predictor on average [109]. Ensemble learning techniques are
often employed when achieving the best performance in a predictive modeling task. The approach is inspired by how
humans explore multiple options before making a final decision. Ensemble design involves two key steps: First, a set of
base learners is generated using the training data; then, one or more of these models is combined to create a unified
prediction model. A diagrammatic representation of an ensemble model for software reliability prediction is illustrated in
Figure 6. Initially, the complete software failure dataset is used to train each classifier. Once the training is complete, each
testing example is presented to all the classifiers to generate their respective outputs. These outputs are then weighted,
and the resulting values are passed through the activation function for further processing.
The process of ensemble learning involves three key stages: ensemble generation, pruning, and integration. During
ensemble generation, a collection of base models is created using a specific base learning algorithm. In the pruning
stage, some of the generated models are eliminated to optimize the ensemble and improve its overall performance.

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

In the literature, there are two types of ensemble techniques: a) homogeneous ensembles—This technique refers to
bagging, boosting, negative correlation, and random correlation; and b) heterogeneous ensembles—This technique
refers to the combination of various ensemble learning techniques.
Ensembled SRP refers to employing multiple estimation techniques (base learners) within an ensemble, guided
by a specific combination rule. This ensemble’s estimate is based on a combination rule that integrates the effort
estimation values provided by the base learners within the ensemble [110]. For software reliability prediction, Zheng
[111] introduced neural network ensembles and tested the method on two software datasets, comparing the results
with those of a single neural network model and various statistical models. The findings of these experiments suggest
that neural network ensembles are more capable of forecasting. Numerous researchers have reported that ensemble
techniques achieve higher prediction accuracy than single techniques. Ardabili et al. [112] provided an excellent
review of software reliability prediction techniques. Mosquera and Hurtado proposed an ensemble model using KNN
and SVM techniques, which achieved a prediction accuracy that surpassed the average accuracy of all individual
base models [113]. Kiran and Ravi [114] presented a combination of linear and nonlinear ensemble techniques and
reported that their approach yields better software reliability prediction accuracy than the existing models. For soft-
ware reliability prediction, many statistical and intelligence approaches such as MLR, MARS, and BPNN, are utilized.
Bal et al. [115] introduced a three-layer architecture for an ensemble model used in software reliability prediction. It
comprises an input layer, a component layer that includes a feed-forward neural network with radial basis functions,
and an output layer that combines the outputs of all component layers through averaging.
Ensemble techniques are classified as linear or nonlinear based on the combination rule used to integrate the
outputs from different learning methods. In a linear ensemble, the outputs are merged using rules based on linear
combinations, such as weighted averages or simple averages [116]. The outputs of a nonlinear ensemble are com-
bined using nonlinear algorithms such as the RBF, MLP, SVR, and decision tree methods. Li et al. [117] applied a selec-
tive ensemble of individual networks using the K-means clustering algorithm for software reliability prediction. The
prediction accuracy of the proposed method was found to be better than that of the BP algorithm. Ali et al. [118] pro-
posed a two-stage prediction approach for detecting defective modules. In the first stage, four supervised machine
learning algorithms are applied, and in the second stage, the predictive accuracy of each classifier is combined into
a voting ensemble to generate the final predictions. The results showed that the proposed method outperformed
other traditional techniques for software defect prediction.
Genetic algorithms (GAs) can estimate optimal parameters in the learning process by leveraging past software
failure data [119]. While the proposed models are created using linear ensembles, it is also necessary to construct
nonlinear models for performance comparison. Elish et al. [120] evaluated different heterogeneous and homogeneous
ensembles of certain optimized hybrid intelligent systems for the SRP. Various nonlinear and linear combiners were
tested. The experimental results revealed that the ensemble had a markedly grater prediction accuracy than the base
method in most cases. Usually, the software failure dataset is very small. A method for enhancing the training dataset
of software failure by using virtual data points was proposed by Behera et al [121]. Following dataset modification,
the ANN, ridge regressor, Bayesian ridge, and KNN models are used to predict software reliability. While the ridge
regressor and the Bayesian ridge regressor perform better on actual data points, the ANN and KNN perform better
when virtual data points are added.
Recently, Gupta et al. [122] proposed an ensemble method for predicting software faults. This study involved the
evaluation of contemporary classifiers such as naive Bayes, and decision tables, as well as ensemble classifiers (AdaBoost,
gradient boosting) for software fault prediction (SFP). The experiments were conducted using three iterations of the
Eclipse datasets. The findings indicate that ensemble methodologies have the potential to enhance the classification
performance in SFP, with XGBoost yielding the most favorable outcomes. Balaram and Vasundra [123] developed an
ensemble-random forest innovative sampling technique to determine the most effective sample representation for
capturing the entire dataset. Feature extraction is accomplished using the bat-induced butterfly optimization (BBO)
method. The effectiveness of the proposed technique was evaluated using metrics such as fault detection accuracy, true
positive rate, and Pearson’s correlation coefficient. The outcomes demonstrate that this technique surpasses alternative
classifiers in terms of performance. Borandag et al. [124] proposed an ensemble ML learning technique comprising
CNN, LSTM, and BiLSTM for software failure prediction that outperforms individual machine learning (ML) techniques
on datasets with large sample sizes.
Aftab et al. [125] developed a novel cloud-based software defect prediction model that leverages machine-learning
fusion techniques at both the data and decision levels. The system uses a two-step prediction approach to identify
problematic modules. In the initial prediction stage, the naive Bayes, ANN, and decision tree techniques were employed.

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

Table 4  Performance of References Algorithms Data set AE/RE MRE NRMSE

ensembled-based software
reliability prediction Zheng [111] Ensemble Musa data set 1.8974
techniques
Kiran & Ravi [114] NEWBPNN Musa data set DS1-DS2 0.1307
Bal et al. [115] Ensemble Musa data set 1.1519
Li et al. [117] KNN Ensemble dynamic flight program 0.7448
Aljahdali & El-Telbany [119] Ensemble Military data set 3.38E-06

Fig. 7  Basic structure of DNN-

based prediction model

The accuracy of these categorization techniques was iteratively improved until it reached its maximum level. The final
prediction was made in the second stage by combining the classifiers with fuzzy logic.
Table 4 compares the performance of ensembled-based software reliability prediction techniques. From the table, it
is observed that different algorithms used different performance measures.
Ensemble methods like Random Forest or Voting Classifiers aggregate predictions from multiple base models, reducing
bias and variance. This robustness makes them inherently more resilient to optimization pitfalls compared to standalone
models. Ensemble techniques are powerful tools for enhancing model performance and robustness, making them suit-
able for software reliability prediction. However, their complexity, computational requirements, and interpretability
challenges necessitate careful design and use. Optimization of computational efficiency can be achieved through parallel
processing. To understand ensemble decisions, model-agnostic interpretation tools can be used.

3.5 Deep learning‑based software reliability prediction techniques

In a feed-forward network, training samples are not correlated, which is important given the highly variable nature of
software failure datasets. To handle complex software failure datasets, a deep neural network can be used [126]. As a
subset of machine learning, deep neural networks employ multilayered neural architectures to mimic the complex
decision-making processes of the human brain. Figure 7 provides a diagrammatic representation of a deep neural net-
work model used for software reliability prediction.
In a feedforward neural network, the input data are not correlated. However, due to its feedback mechanisms, a recur-
rent neural network (RNN) is better suited for handling time-varying data, resulting in more accurate predictions. The
literature presents various recurrent structures designed for software reliability prediction. Deep neural networks, such
as RNNs and long short-term memory (LSTM), have been employed to forecast the reliability of software in the context
of time series data. Wang and Zhang [127] proposed a deep neural network for software failure prediction. To measure

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

software reliability and quantify the number of failures, they employed a recurrent neural network encoder-decoder.
Recently, Mohana et al. [128] conducted an exploratory analysis to evaluate the effectiveness of well-known methods,
such as logistic regression, decision trees, the random forest algorithm, SVM algorithms, and deep neural networks
(DNNs).
A comprehensive comparative analysis of deep learning techniques applied to both file-level and change-level soft-
ware defect prediction (SDP) was carried out [129]. The study also highlights the difficulties and concerns related to SDP
when employing deep learning approaches. Karunanithi et al. [130] used an FNN and a recurrent network to forecast
software reliability. They compared their findings to 14 distinct datasets and found that the NN achieved greater predic-
tive accuracy at end-point predictions than did analytical models. For the predictability metric, the authors included
only the average error, average error (AE), average bias (AB), normalized average error (NAE), and normalized root mean
square error (NRMSE). Hu et al. [131] conducted their research in two stages. Mostly in the initial phase, recurrent neural
networks are developed to predict fault diagnosis and correction together. In the second stage, the genetic algorithm
is utilized to develop various designs for accurate predictions. According to the findings of a real-world investigation,
ANN models outperform analytical models. Roy et al. [132] proposed a recurrent neural network structure that is entirely
based on feedforward neural networks to evaluate software reliability by learning the dynamic behavior of software
failure data. Their findings indicate that the proposed models offer greater accuracy in software reliability prediction
than traditional ANN-based models.
Bhuyan et al. [133] conducted comparative research on predictive modeling of software reliability and developed an
improved fuzzy min-max technique integrated with recurrent neural networks. Using a recurrent neural network with
back-propagation over time methods, Bhuyan et al. [134] developed a novel method for forecasting software reliability,
which gave more accurate results than prior models. Furthermore, Zemouri and Patic [135] used an evolutionary connec-
tionist method to provide a recurrent RBFN for the SRP. Fan et al. [136] introduced a deep learning-based technique called
DP-ARNN to identify probable code errors in software, in which DP-ARNN uses an RNN to automatically create syntactic
and semantic characteristics from source code, outperforming state-of-the-art baseline methods by fourteen percent.
Benaddy et al. [137] used software failure data to train a recurrent neural network (RNN) to forecast software failures
and it outperformed the existing models. In comparison to well-known prediction algorithms, Tian and Noor [138]
claimed that their proposed model yielded a reduced average relative prediction error. A FLANN model based on recur-
rent chemical reaction optimization was presented by Behera et al. [139] to forecast software reliability. To predict soft-
ware reliability, they used recurrent FLANN, and CRO was used to estimate the model’s parameters. The appropriate
distribution of input data was achieved using a time-series technique with logarithmic scaling.
Researchers have proposed various predictive approaches using traditional deep learning techniques, but a common
issue among them is the high rate of misclassification. Adjusting the correct hyperparameters is a challenging task in
deep learning. To address these issues, Pemmada et al. [140] proposed a deep neural network with a memetic firefly algo-
rithm for improved software risk prediction. Most of the existing research on SDP has been conducted without feature
extraction for high-dimensional data, leading to decreased generalizability. Recently, Zhang et al. [141] proposed a deep
Q-learning-based feature extraction model for efficient software defect prediction. Table 5 compares the performance
of DL-based software reliability prediction techniques. From the table, it is observed that the performance measures
for different techniques are either MSE or MSE. It is also found that the AE value of the DNN technique is 0.07 which is
better than other techniques.
DL excels in processing high-dimensional and unstructured data such as images, text, and audio, making it invalu-
able for fields like computer vision, NLP, speech recognition, and software reliability prediction. Deep learning is a
transformative technology that has revolutionized numerous fields. However, its effectiveness comes at the cost of

Table 5  Performance of References Algorithms Data set MSE AE/RE

DL-based software reliability
prediction techniques Wang & Zhang [127] RNN encoder-decoder Musa data set 0.07
Karunanithi & Whitley [130] FFN Musa data set 2.92
Roy et al. [132] RNN-based dynamic weighted Musa data set 0.1069
combination model
Bhuyan et al. [134] Feed Forward Back-propagation Musa, Iyer, and Lee 3.0018
3.3540
Benaddy et al. [137] RNN Musa data set 2.6617

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

high data and computational requirements, interpretability issues, and ethical challenges. For data requirements,
data augmentation and transfer learning can be employed to mitigate data scarcity. Cloud services and specialized
hardware can be utilized to meet computational demand.

3.6 Hybridized techniques

Numerous CI models such as NN, FL, GA, and PR have been successfully developed to achieve more accurate results
[16, 142, 143]. However, no single CI technique is preferable to the other techniques in every situation. Therefore, the
current trend involves developing hybrid techniques, that can eliminate the weaknesses of individual techniques
and capitalize on their strengths [144]. In comparison to the single CI approach, hybrid strategies produce excellent
outcomes. A Venn diagram of the hybridization techniques used is illustrated in Figure 8.
Pati and Sukla [145] proposed a hybrid ARIMA (ARIMA+NN) model for the prediction of software reliability based
on real-life data on software failures. The empirical result showed that the proposed hybrid model improved the pre-
diction accuracy. Malhotra and Negi [146] proposed a software reliability model using the PSO algorithm for SRGM
parameter estimation and subsequently compared the results with those of the GA algorithm. The hybrid technique
(ANN-PSO) has been demonstrated to be more predictive of software reliability [147]. The Log function is used to
increase the input values in a unique ANN-based technique with an extra layer within the hidden and input layers.
In their work, Al Gargoor and Saleem demonstrated an ANN-PSO-based model through three software failure data-
sets [148]. From the results, it was concluded that the proposed model was more applicable for software reliability
prediction. Yang et al. [149] proposed a hybrid algorithm for estimating model parameters in software defect predic-
tion, combining particle swarm optimization (PSO) with the sparrow search algorithm (SSA). They developed a new
fitness function based on the maximum likelihood estimation of parameters and utilized it for parameter initializa-
tion. The experimental results showed that the hybrid SSA-PSO algorithm has a better solution, convergence speed,
and stability than the conventional model. Usually, software failure datasets are very large with high dimensions
which requires more classification time. To address this issue, Qin [150] used principal component analysis (PCA) for
dimension reduction and subsequently applied the PSO-SVM model for software reliability prediction. The results
demonstrated better predictive capability in less time.
For parameter optimization of a predictive model, many swarm intelligence algorithms have been widely used in
the literature. Each optimization algorithm has advantages and disadvantages that impact the predictive capabil-
ity of the model. Zen et al. [151] proposed a hybrid software reliability model in which the model parameters are
estimated using the wolf pack algorithm (WPA-PSO). The simulation results show better prediction accuracy than
that of a single algorithm.

Fig. 8  Hybridization tech-

niques for SRP

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

In recent years, neuro-fuzzy systems, which integrate neural networks’limited learning and processing capabilities
with fuzzy systems’high-level human-like reasoning powers, have been used for prediction. In the field of fuzzy logic
reliability prediction, several research publications have been published [152, 153].
Yuan et al. [154] used a fuzzy-neural hybrid network to develop a method for evaluating software reliability. The
authors tested the network and discovered that the ANFIS has faster learning than the radial basis function neural net-
work. For software reliability prediction, Zhao et al. [155] used a fuzzy wavelet neural network. The results of the experi-
ments reveal that the proposed technique can quickly estimate the network architecture based on software failure data.
Shanmugam et al. [156] explored the analysis and enhancement of various ant colony optimization (ACA) methods for
software reliability models, focusing on their capacity to anticipate fault occurrences during the development phase. A
robust testing procedure was also explored, to aid in the precise scheduling of software releases and the effective allo-
cation of project resources. Employing a novel approach, Babaie [157] quantified software quality parameters through
the utilization of a neuro-fuzzy model. The experimental investigation demonstrated that the proposed method yields
more precise software reliability measurements than both the fuzzy multi-criteria and fuzzy approaches. This research
further involved assessing the estimation accuracy of these methods. Here we briefly describe relevant work on software
reliability prediction in various phases using various methodologies. Table 6 compares the performance of hybrid tech-
niques for software reliability prediction. From the table, it is observed that the performance of most of the techniques is
based on NRMSE. Among all techniques, the NRMSE value of the RANFIS-MDE technique is 0.1106 which is the minimum
among other techniques.
Hybrid techniques offer a powerful approach to tackling complex and multifaceted problems by combining the
strengths of multiple methods. However, they come with added complexity, computational demands, and integration
challenges. Design modular architectures that allow for easier integration. Employ automated machine learning tools
for parameter tuning.
The taxonomy presented in Figure 9 is formulated based on recent nonparametric software reliability modeling tech-
niques, emphasizing the connections and associations among these studies. In the figure, the recent research works on
SRP and the relationship between these works is highlighted.
Computational intelligence approaches have become the most popular and commonly utilized technique in the
field of software engineering because they provide benefits such as reduced cost, time, and maintenance, as well as
enhanced productivity. For many software developers and engineers, automated technologies in software engineering
have become the de-facto standard. The methods of CI techniques for software reliability prediction are compared in
Table 7.

4 Critical analysis and investigation

This section presents an overview of the findings from the articles reviewed from 2005 to 2024. A concise analysis was
carried out to illustrate the effectiveness of all the methods under consideration. The results are presented in tables,
showcasing the evidence for each method. Table 7 lists the datasets utilized in 40 studies. The table indicates the
utilization of 25 datasets specifically for software reliability prediction. Various computational intelligence techniques
have been employed to predict software reliability. Several studies involve hybrid techniques for such predictions,
subsequently comparing them to determine the most effective technique. Intelligent systems have been generated
by modeling biological and natural intelligence. These intelligent algorithms, which are termed CIs, include ANNs,
evolutionary computing, SI, ensembles, and FSs. Among the diverse studies on intelligent methods depicted in

Table 6  Performance of References Algorithms Data set MSE MRE NRMSE

hybrid software reliability
prediction techniques Pati & Sukla [145] ARIMA + ANN Real-time command control 0.480871
Malhotra & Negi [146] PSO Military 11.4541
Bisi & Goyal [147] ANN + PSO Musa data set 0.1283
Behera et al. [152] RANFIS + MDE Musa data set 0.1106
Roy et al. [159] BPNN Musa data set 0.1449
Raamesh et al. [162] BSO + LAHC Bugzilla and Firefox 1.92

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

Fig. 9  Schematic representa-

tion of related works using
intelligent systems

Figure 10, ANN (encompassing MLP, and BP) is the most commonly employed technique (mentioned in 117 studies
constituting [42% of all studies]), followed by SEC which is (used in 61 studies [22% of all studies]).
When it comes to software reliability prediction, a variety of machine learning (ML) algorithms have been applied.
Each technique has its strengths, but some have shown more consistent and superior performance across studies
and applications. It is observed that the Ensemble methods dominate both usage and performance due to gener-
alization strength and low overfitting risk. Deep learning is increasingly popular, especially for large and complex
datasets whereas Hybrid models (e.g., Neuro-Fuzzy) often achieve very high accuracy, but are used less often due
to their implementation difficulty. Traditional techniques like ANNs still widely cited, especially in simpler datasets.
Figure 11 presents a comparative analysis of mostly utilized algorithms for performing the better results in software
reliability prediction.
SRP employs numerous accuracy measures for performance analysis. Table 8 displays the accuracy measures uti-
lized in the literature, and Figure 12 presents a comparative analysis of t-values. Among these measures, the NRMSE
is the most frequently used, appearing in 46 studies (accounting for 57.5% of all studies). Several techniques are
amalgamated to create hybrid models, aiming for a more precise prediction of software reliability. We have also ana-
lysed the performance measure of the different SRP methods. It was found that the hybrid methods outperformed
other techniques.
Figure 13 presents the hybrid models employed in software reliability prediction. The most frequently utilized com-
bined technique was ANN-FLS, mentioned in 6 studies (constituting 17% of all studies).
Based on the current scenario, software reliability prediction methods should incorporate (i) an intelligent system,
(ii) automated prediction mechanisms, and (iii) early detection of software reliability issues during the development
stage. These elements establish a protective framework, enhancing the effectiveness of software reliability predictions.
Intelligent systems have become a valuable tool for predicting software reliability, but significant challenges remain.
High-quality historical failure data is essential for training accurate models, yet it is often scarce, especially for new pro-
jects or proprietary software. Moreover, software failure data tends to be highly imbalanced, as failures are relatively
rare compared to normal operations, making it difficult for models to accurately capture these patterns. Choosing and
tuning the right model type (e.g., neural networks, decision trees) is critical, but it is also time-consuming and requires
computational resources. Overfitting is a risk, especially with complex models trained on limited data, which reduces the
ability to generalize well on new data. There is no universal standard for assessing reliability prediction models, making
it challenging to compare models across studies. Additionally, benchmark datasets are limited, and validation can be
complex, as the impact of reliability issues often only becomes clear over extended use. In this extensive survey, though
we have systematically analyzed several research papers on SRP models, we have not compared the traditional statistical
model with the current machine learning models.

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Table 7  Comparison of software reliability prediction techniques using CIs
Sl. No Refer- Model Dataset Advantages/Limitations
Review

ences

1 Ref. [158] NN + Simulated Annealing Military, Real-Time Command & Control, and Operating System Greater prediction accuracy of software cumulative failure

Vol:.(1234567890)
datasets obtained from Bell Laboratories
2 Ref. [159] ANN + GA DS1, DS2, DS3 Better software reliability prediction than other NHPP model
3 Ref. [160] Nonhomogenous poisson process Real-world software applications data NHPP model gives better results in software reliability
4 Ref. [161] MCSA + ABC Banking application software Reducing the failure rate
5 Ref. [162] Brainstorm Optimization + LSTM Bugzilla and Firefox datasets, Demonstrated superior results with less execution time
6 Ref. [163] NN + FL 47 classroom-based projects data ANFIS improves the reliability evaluation
7 Ref. [164] ANN + fuzzy PSO DS1, DS2, DS3 Better fitting performance than standard PSO
8 Ref. [165] RGA​ Musa dataset Real value genetic algorithm is more effective in software reli-
ability growth model
9 Ref. [166] ARIMA Musa datasets Deliver precise predictions
10 Ref. [167] IT2 FLS + ANN 28 real software project data The model showed better predictive accuracy in terms of perfor-
mance metrics
Discover Computing

11 Ref. [168] RNN + BPTT DS1, DS2, DS3 The RNN demonstrates precise behavior in predicting reliability
12 Ref. [169] GA-ANN Eclipse dataset SMOTE-Tomek data sampling technique to manage imbalanced
data
13 Ref. [170] Deep Learning 8 open-source project datasets Bugzilla data AverageF1-score of the proposed model surpassed other tech-
niques by 16.88%
14 Ref. [171] ARIMA-ANN DS1, DS2, DS3 The proposed model enhanced decision-making and heightened
(2025) 28:90

customer satisfaction
15 Ref. [172] NSGAII-kNN CM1, KC1, KC3, MC1, AR3, PC4 NSGA-II optimizes the parameters in the KNN classifier
16 Ref. [173] GP + Boosting Failure datasets are collected from a program Space The proposed algorithm’s computational cost is 300 times lower
than the traditional model
18 Ref. [174] Differential Equation Real-time command and control system, telecommunication Better than many probability-base SRGM
system
19 Ref. [175] NHMP-based SRM DS1-DS20 datasets Made optimistic estimation of software reliability
20 Ref. [176] RNN + WSE DS1-DS6 Showed higher prediction accuracy than the conventional
method
21 Ref. [177] CGWO Tandem DS1, DS2 datasets The process is fully mechanized and does not require any input or
intervention from the user
22 Ref. [178] SDE-based SRGM, IT2 FN Mozilla dataset Offered a more practical approach to determining the software’s
release time

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

23 Ref. [179] SSA + SVR The dataset collected from the maintenance stage of Boeing The proposed model outperforms the traditional model and
737 aircraft could be used as a tool for reliability forecast applications
24 Ref. [180] ANN + Adam optimizer JM1 dataset Better prediction accuracy
25 Ref. [181] NFS-DELI algorithm DS1, DS2, DS3 The NFS-DELI method gives less RMSE than another optimization
algorithm
26 Ref. [182] ANN + BP DS1, DS2 Better results
| https://doi.org/10.1007/s10791-025-09597-z
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

Fig. 10  Summarized report of publications on intelligent systems used for the SRP

Fig. 11  Utilization of ML algo-

rithms for SRP Fuzzy-Logic

SEC Ensembled Ensembled

ANN
Deep Learning
Hybrid
Hybrid
SEC
Fuzzy-Logic
Deep Learning ANN

5 Conclusion

This survey presents a comprehensive analysis of 140 studies on software reliability prediction conducted over the
past two decades, focusing on intelligent machine-learning techniques. This review is significant because it contrib-
utes to the ongoing research on enhancing methods for precise software reliability prediction. This approach will
ultimately impact the effectiveness and applicability of software systems in modern society. Our review highlights
the methodologies using advanced machine learning and hybrid models, illustrating the growing software complex-
ity and reliability demands. The survey reveals several key trends and findings. Machine learning techniques have
gained significant popularity due to their capability to handle complex, high-dimensional data. Additionally, the
application of hybrid models has enhanced the ability to identify key reliability indicators, leading to more robust
predictive models.

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

Table 8  Performance Performance Description Equation

measures used Measures

MSE Mean Square Error 1

n � �2
yi − y�i
∑
n
i=1

AE/RE Average Error 1 ∑n

RE = abs((yi − yi� )∕yi ), AE = n i=1 REi
MAE Mean Absolute Error n �� � �
abs yi − y�i ∕n
∑
i=1

MRE Magnitude Relative Error 1

n �� � �
abs yi − y�i ∕yi
∑
MAPE = n
i=1

RMSE Root Mean Square Error

�
n � �2
1
yi − y�i
∑
RMSE = n
i=1

NRMSE Normalized Root Mean Square Error

�∑ 2
n
i=1 (yi −y�i )
NRMSE = ∑n 2
i=1 yi

Here, yi is the actual time, and y′i the estimated time between two successive failures and n is the number
of observations

Fig. 12  Evaluation factors in

various studies

Despite these advances, the challenges remain including the lack of standardized datasets, issues of interpret-
ability, and computational cost of training complex models. These issues highlight potential areas for future research,
particularly in the exploration of explainable AI and the development of benchmarking standards. This survey aims
to serve as a foundational reference for researchers and practitioners, encouraging further innovations and practical
adaptation of intelligent systems in software reliability prediction.

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

Fig. 13  Hybrid techniques

used in various studies of soft-
ware reliability prediction

Author contributions Ajit Kumar Behera wrote the main manuscript. Pamela Chaudhury prepared all figures. Ch. Sanjeev Kumar Dash worte
the reference section. All authors reviewed the manuscript.

Funding The authors received no financial support for the research, authorship, and/or publication of this article.

Data availability No datasets were generated or analysed during the current study.

Declarations
Ethics approval and consent to participate Not applicable

Consent for publication Not applicable

Competing interests The authors declare no competing interests.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which
permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to
the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You
do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party
material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds
the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​
mmons.​org/​licen​ses/​by-​nc-​nd/4.​0/.

References
1. Johnson J, Mulder H. 2020. Endless modernization. Technical report, The Standish Group International, Incorporated.
2. https://​www.​softw​arete​sting​news.​co.​uk/
3. Musa JD. 1979. Software reliability data. Technical report, data and analysis center for software, Rome Air Development Center, Griffis
AFB.
4. Pai PF, Hong WC. Software reliability forecasting by support vector machines with simulated annealing algorithms. J Syst Softw.
2006;79(6):747–55.
5. Dalal SR, Lyu MR, Mallows CL. 2014. Software reliability. Bellcore Luscent Technologies, AT&T Research.
6. Ramani S, Gokhale SS, Trivedi KS. SREPT: software reliability estimation and prediction tool. Perform Eval. 2000;39(1–4):37–60.
7. Musa JD. A theory of software reliability and its application. IEEE Trans Software Eng. 1975;1(03):312–27.
8. Su YS, Huang CY. Neural-network-based approaches for software reliability estimation using dynamic weighted combinational models.
J Syst Softw. 2007;80(4):606–15.
9. van Driel WD, Bikker JW, Tijink M. Prediction of software reliability. Microelectron Reliab. 2021;119: 114074.
10. Diwaker C, Tomar P, Solanki A, Nayyar A, Jhanjhi NZ, Abdullah A, Supramaniam M. A new model for predicting component-based software
reliability using soft computing. IEEE Access. 2019;7:147191–203.
11. Benala TR, Dehuri S, Mall R. Computational intelligence in software cost estimation: An Emerging Paradigm. ACM SIGSOFT Softw Eng.
2012;37(3):1–7.

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

12. Engelbrecht AP. Computational intelligence: an introduction. Hoboken: John Wiley & Sons; 2007.
13. Yang T. Performance analysis in the reliability attributes of nhpp software reliability attributes of NHPP Software reliability model apply-
ing exponential and inverse-exponential lifetime distribution. J Theo Appl Inf Technol. 2022;100(22):3.
14. Rastogi N, Rastogi S, Darbari M. Survey on software reliability prediction using soft computing. Int J Comput Eng Technol. 2018;9(4):1.
15. Oveisi S, Moeini A, Mirzaei S, Farsi MA. Software reliability prediction: a survey. Qual Reliab Eng Int. 2023;39(1):412–53.
16. Kong LS, Jasser MB, Ajibade SSM, Mohamed AW. A systematic review on software reliability prediction via swarm intelligence algorithms.
J King Saud Univ-Comput Inf Sci. 2024;36:102132.
17. Rana R, Staron M, Berger C, Hansson J, Nilsson M, Törner F, Höglund C. Selecting software reliability growth models and improving
their predictive accuracy using historical project data. J Syst Softw. 2014;98:59–78.
18. Lakshmanan I, Ramasamy S. An artificial neural network approach to software reliability growth modeling. Proc Comput Sci.
2015;57:695–702.
19. Kapur PK, Pham H, Anand S, Yadav K. A unified approach for developing software reliability growth models in the presence of
imperfect debugging and error generation. IEEE Trans Reliab. 2011;60(1):331–40.
20. Sinha S, Goyal NK, Mall R. Early prediction of reliability and availability of combined hardware-software systems based on functional
failures. J Syst Architect. 2019;92:23–38.
21. Rathi SC, Misra S, Colomo-Palacios R, Adarsh R, Neti LBM, Kumar L. Empirical evaluation of the performance of data sampling and
feature selection techniques for software fault prediction. Expert Syst Appl. 2023;223: 119806.
22. Pandey AK, Goyal NK, Pandey AK, Goyal NK. 2013. Multistage model for residual fault prediction. Early software reliability prediction:
a fuzzy logic approach, 59–80.
23. Samal U, Kumar A. Redefining software reliability modeling: embracing fault-dependency, imperfect removal, and maximum fault
considerations. Qual Eng. 2023;36(3):500–9. https://​doi.​org/​10.​1080/​08982​112.​2023.​22410​67.
24. Samal U, Kumar A. A software reliability model incorporating fault removal efficiency and its release policy. Comput Stat.
2024;39(6):3137–55.
25. Pandey AK, Goyal NK. A fuzzy model for early software fault prediction using process maturity and software metrics. Int J Electron
Eng. 2009;1(2):239–45.
26. Samal U, Kumar A. Empowering software reliability: Leveraging efficient fault detection and removal efficiency. Quality Eng.
2024;94:1–12.
27. Santhosh S, Khatter K, Relan D. 2023. Software fault prediction using particle swarm optimization and random forest. In Proceedings of
International Conference on Data Science and Applications: ICDSA 2022, Volume 1 (pp. 843–851). Singapore: Springer Nature Singapore.
28. Rajput PK, Pal R. 2023. Genetic algorithm-based clustering with neural network classification for software fault prediction. In Proceed-
ings of International Conference on Data Science and Applications: ICDSA 2022, Volume 1 (pp. 399–414). Singapore: Springer Nature
Singapore.
29. Rao KN, Reddy CS. A novel under sampling strategy for efficient software defect analysis of skewed distributed data. Evol Syst.
2020;11(1):119–31.
30. Gokhale SS, Trivedi KS. Analytical models for architecture-based software reliability prediction: a unification framework. IEEE Trans Reliab.
2006;55(4):578–90.
31. Goševa-Popstojanova K, Trivedi KS. Architecture-based approach to reliability assessment of software systems. Perform Eval.
2001;45(2–3):179–204.
32. Kuo L, Yang TY. Bayesian computation for nonhomogeneous Poisson processes in software reliability. J Am Stat Assoc.
1996;91(434):763–73.
33. Littlewood B, Verrall JL. A Bayesian reliability growth model for computer software. J Roy Stat Soc Ser C. 1973;22(3):332–46.
34. Littlewood B, Salako K, Strigini L, Zhao X. On reliability assessment when a software-based system is replaced by a thought-to-be-better
one. Reliab Eng Syst Saf. 2020;197: 106752.
35. Bai CG. Bayesian network-based software reliability prediction with an operational profile. J Syst Softw. 2005;77(2):103–12.
36. Kumar P, Singh LK, Kumar C. Suitability analysis of software reliability models for its applicability on NPP systems. Qual Reliab Eng Int.
2018;34(8):1491–509.
37. Torrado N, Wiper MP, Lillo RE. Software reliability modeling with software metrics data via Gaussian processes. IEEE Trans Softw Eng.
2012;39(8):1179–86.
38. Chatterjee S, Maji B. A Mahalanobis distance-based algorithm for assigning rank to the predicted fault-prone software modules. Appl
Soft Comput. 2018;70:764–72.
39. Singh Y, Kaur A, Malhotra R. Empirical validation of object-oriented metrics for predicting fault proneness models. Software Qual J.
2010;18:3–35.
40. Lyu MR. 2007. Software reliability engineering: a roadmap. In Future of Software Engineering (FOSE’07) (pp. 153–170). IEEE.
41. Wang J, Wu Z, Shu Y, Zhang Z. An optimized method for software reliability model based on a nonhomogeneous Poisson process. Appl
Math Model. 2016;40(13–14):6324–39.
42. Aljahdali SH, Sheta A, Rine D. 2001. Prediction of software reliability: a comparison between regression and neural network non-para-
metric models. In Proceedings ACS/IEEE International Conference on Computer Systems and Applications (pp. 470–473). IEEE.
43. Okutan A, Yildiz OT. Software defect prediction using Bayesian networks. Empir Softw Eng. 2014;19(1):154–81.
44. Kiran NR, Ravi V. 2007. Software reliability prediction using wavelet neural networks. In International Conference on Computational
Intelligence and Multimedia Applications (ICCIMA 2007) (Vol. 1, pp. 195–199). IEEE.
45. Jungang L, Yunliang J, Qing S, Zhangguo S, Zhen W, Ruiqin W. Software reliability prediction via relevance vector regression. Neurocom-
puting. 2016;186:66–73.
46. Aljahdali SH, Buragga KA. Employing four ANNs paradigms for software reliability prediction: an analytical study. ICGST Int J Artif Intell
Mach Learn. 2008;8(2):1–8.
47. Goyal S, Bhatia PK. Comparison of machine learning techniques for software quality prediction. Int J Knowl Syst Sci. 2020;11(2):20–40.

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

48. Liapis CM, Karanikola A, Kotsiantis S. Data-efficient software defect prediction: a comparative analysis of active learning-enhanced
models and voting ensembles. Inf Sci. 2024. https://​doi.​org/​10.​1016/j.​ins.​2024.​120786.
49. Song Q, Guo Y, Shepperd M. A comprehensive investigation of the role of imbalanced learning for software defect prediction. IEEE Trans
Softw Eng. 2018;45(12):1253–69.
50. Son LH, Pritam N, Khari M, Kumar R, Phuong PTM, Thong PH. Empirical study of software defect prediction: a systematic mapping. Sym-
metry. 2019;11(2):212.
51. Bhuyan MK, Mohapatra DP, Sethi S. Measures for predicting software reliability using time recurrent neural networks with back-propa-
gation. ACM SIGSOFT Softw Eng Notes. 2015;40(5):1–8.
52. Jaiswal A, Malhotra R. Software reliability prediction using machine learning techniques. Int J Syst Assurance Eng Manage.
2018;9(1):230–44.
53. Bisi M, Goyal NK. Software development efforts prediction using artificial neural network. IET Softw. 2016;10(3):63–71.
54. Karunanithi N, Malaiya YK, Whitley LD. 1991. Prediction of software reliability using neural networks. In ISSRE (pp. 124–130).
55. Hu QP, Dai YS, Xie M, Ng SH. 2006. Early software reliability prediction with extended ANN model. In 30th Annual International Computer
Software and Applications Conference (COMPSAC’06) (Vol. 2, pp. 234–239). IEEE.
56. Noekhah S, Salim NB, Zakaria NH. 2017. Predicting software reliability with a novel neural network approach. In International Conference
of Reliable Information and Communication Technology (pp. 907–916). Springer, Cham.
57. Vishwanath SP. 2006. Software reliability prediction using neural networks (Doctoral dissertation, IIT Kharagpur).
58. Kaur J, Singh S, Kahlon KS, Bassi P. Neural network-a novel technique for software effort estimation. Int J Comput Theory Eng. 2010;2(1):17.
59. Wu CY, Huang CY. A study of incorporation of deep learning into software reliability modeling and assessment. IEEE Trans Reliab.
2021;70(4):1621–40.
60. Cai KY, Cai L, Wang WD, Yu ZY, Zhang D. On the neural network approach in software reliability modeling. J Syst Softw. 2001;58(1):47–62.
61. Rumelhart DE, Hinton GE, Williams RJ. Learning representations by back-propagating errors. Nature. 1986;323(6088):533.
62. Singh Y, Kumar P. 2010. Prediction of software reliability using feed-forward neural networks. In 2010 International Conference on Com-
putational Intelligence and Software Engineering (pp. 1–5). IEEE.
63. Dehuri S, Cho SB. A comprehensive survey on functional link neural networks and an adaptive PSO–BP learning for CFLNN. Neural
Comput Appl. 2010;19(2):187–205.
64. Arya A, Malik SK. Software fault prediction using K-mean-based machine learning approach. Int J Perform Eng. 2023;19(2):133.
65. Li Z. 2020. Optimization Strategy of BP Neural Network for Software Reliability Prediction. In 2020 International Conference on Intelligent
Transportation, Big Data & Smart City (ICITBS) (pp. 897–900). IEEE.
66. Tian L, Noore A. Evolutionary neural network modeling for software cumulative failure time prediction. Reliab Eng Syst Saf.
2005;87(1):45–51.
67. Kumar P, Singh Y. An empirical study of software reliability prediction using machine learning techniques. Int J Syst Assurance Eng Man-
age. 2012;3(3):194–208.
68. Rathore SS, Kumar S. Towards an ensemble-based system for predicting the number of software faults. Expert Syst Appl. 2017;82:357–82.
69. Bisi M, Goyal NK. Artificial neural network applications for software reliability prediction. Performability engineering series. Hoboken:
John Wiley & Sons; 2017.
70. Lo JH. 2009. The implementation of artificial neural networks applying to software reliability modeling. In 2009 Chinese control and
decision conference (pp. 4349–4354). IEEE.
71. Jin C. Software reliability prediction based on support vector regression using a hybrid genetic algorithm and simulated annealing
algorithm. Instit Eng Technol. 2011;5(4):398–405.
72. Tian L, Noore A. Online prediction of software reliability using an evolutionary connectionist model. J Syst Softw. 2005;77(2):173–80.
73. Bal PR, Mohapatra DP. 2017. Software reliability prediction based on radial basis function neural network. In Advances in Computational
Intelligence (pp. 101–110). Springer, Singapore.
74. Rath SK, Sahu M, Das SP, Mohapatra SK. 2022. Hybrid Software reliability prediction model using feature selection and support vector
classifier. In 2022 International Conference on Emerging Smart Computing and Informatics (ESCI) (pp. 1–4). IEEE.
75. Ndahi BP, Abisoye OA, Aliyu HO, Ojerinde OA. Development of pi sigma neural network model for the prediction of software reliability
using 5 NASA failure datasets. East Afr J Inf Technol. 2023;6(1):135–48.
76. Wang K, Liu L, Yuan C, Wang Z. Software defect prediction model based on LASSO–SVM. Neural Comput Appl. 2021;33(14):8249–59.
77. Toofani A, Garg H. 2023. Analysis of feature selections during fault prediction using various ML algorithms. In AIP Conference Proceed-
ings (Vol. 2721, No. 1). AIP Publishing.
78. Rajpurohit J, Sharma TK, AbrahamVaishali A. Glossary of metaheuristic algorithms. Int J Comput Inf Syst Ind Manage Appl. 2017;9:181–205.
79. Roy P, Mahapatra GS, Dey KN. An efficient particle swarm optimization-based neural network approach for software reliability assess-
ment. Int J Reliab Qual Saf Eng. 2017;24(04):1750019.
80. Bisi M. 2020. ANN-PSO approach on logistic model based software reliability prediction. In 2020 IEEE 4th Conference on Information &
Communication Technology (CICT) (pp. 1–4). IEEE.
81. Costa EO, Vergilio SR, Pozo A, Souza G. 2005. Modeling software reliability growth with genetic programming. In 16th IEEE International
Symposium on Software Reliability Engineering (ISSRE’05) (pp. 10-pp). IEEE.
82. Zhang Y, Chen H. 2006. Predicting for MTBF failure data series of software reliability by genetic programming algorithm. In Sixth Inter-
national Conference on Intelligent Systems Design and Applications (Vol. 1, pp. 666–670). IEEE.
83. Aljahdali SH, El-Telbany ME. 2009. Software reliability prediction using multi-objective genetic algorithm. In 2009 IEEE/ACS International
Conference on Computer Systems and Applications (pp. 293–300). IEEE.
84. Nayak SC, Misra BB, Behera HS. ACFLN: artificial chemical functional link network for prediction of stock market index. Evol Syst.
2019;10(4):567–92.
85. Lam AY, Li VO. Chemical-reaction-inspired metaheuristic for optimization. IEEE Trans Evol Comput. 2009;14(3):381–99.
86. Behera AK, Nayak SC, Dash CSK, Dehuri S, Panda M. 2019. Improving software reliability prediction accuracy using CRO-based FLANN.
In Innovations in Computer Science and Engineering (pp. 213–220). Springer, Singapore.

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

87. Nayak SC. Bitcoin closing price movement prediction with optimal functional link neural networks. Evolut Intell. 2021;14:1–15.
88. Behera AK, Dash CSK, Panda M, Dehuri S, Mall R. A state-of-the-art neuro-swarm approach for prediction of software reliability. Int J Adv
Intell Paradigm. 2021;20(3–4):296–322.
89. Behera AK, Panda M, Nayak SC, Dash CSK. 2022. An artificial electric field algorithm and artificial neural network-based hybrid model
for software reliability prediction. In computational intelligence in data mining: Proceedings of ICCIDM 2021 (pp. 271–279). Singapore:
Springer Nature Singapore.
90. Rahman MNM, Nugroho RA, Faisal MR, Abadi F, Herteno R. Optimized multi correlation-based feature selection in software defect pre-
diction. TELKOMNIKA Telecommun Comput Electron Control. 2024;22(3):598–605.
91. Wang H, Arasteh B, Arasteh K, Gharehchopogh FS, Rouhi A. A software defect prediction method using binary gray wolf optimizer
and machine learning algorithms. Comput Electr Eng. 2024;118: 109336.
92. Oueslati R, Manita G. 2024. Software defect prediction using integrated logistic regression and fractional chaotic grey wolf optimizer.
In ENASE (pp. 633–640).
93. Malhotra R, Khan K. A novel software defect prediction model using two-phase grey wolf optimisation for feature selection. Cluster
Comput. 2024;27:1–23.
94. Nayak SC, Misra BB, Behera HS. Efficient financial time series prediction with evolutionary virtual data-position exploration. Neural
Comput Appl. 2019;31(2):1053–74.
95. Behera AK, Panda M. 2021. Efficient software reliability prediction with evolutionary virtual data position exploration. in handbook
of research on automated feature engineering and advanced applications in data science (pp. 275–285). IGI Global.
96. Zadeh LA. 1996. Fuzzy sets. In fuzzy sets, fuzzy logic, and fuzzy systems: selected papers by Lotfi A Zadeh (pp. 394–432).
97. Quyoum A, Dar MD, Quadri SMK. Improving software reliability using software engineering approach-a review. Int J Comput Appl.
2010;10(5):41–7.
98. Radjenovic D, Hericko M, Torkar R, Zivkovic A. Software fault prediction metrics: a systematic literature review. Inf Softw Technol.
2013;55(8):1397–418.
99. Sahu K, Alzahrani FA, Srivastava RK, Kumar R. Evaluating the impact of prediction techniques: software reliability perspective. Comput
Mater Contin. 2021;67(2):1471.
100. Kumar R, Khatter K, Kalia A. Measuring software reliability: a fuzzy model. ACM Sigsoft Softw Eng Notes. 2011;36(6):1–6.
101. Chatterjee S, Nigam S, Singh JB, Upadhyaya LN. Application of fuzzy time series in the prediction of time between failures & faults
in software reliability assessment. Fuzzy Inf Eng. 2011;3(3):293–309.
102. Roy P, Mahapatra GS, Dey KN. Forecasting of software reliability using neighborhood fuzzy particle swarm optimization based novel
neural network. IEEE/CAA J Autom Sin. 2019;6(6):1365.
103. Shen Q, Lou J, Zhang X, Jiang Y. Failure prediction by regularized fuzzy learning with intelligent parameters selection. Appl Soft
Comput. 2021;100: 106952.
104. Aljahdali S, Debnath NC. 2004. Improved software reliability prediction through fuzzy logic modeling. In IASSE (pp. 17–21).
105. Yadav HB, Yadav DK. A fuzzy logic based approach for phase-wise software defects prediction using software metrics. Inf Softw
Technol. 2015;63:44–57.
106. Khatatneh, K., & Mustafa, T. (2009). Software reliability modeling using soft computing technique. European Journal of Scientific
Research, 26(1), 147–152. ISSN 1450–216X.
107. Dimov A, Punnekkat S. 2010. Fuzzy reliability model for component-based software systems. In 2010 36th EUROMICRO conference
on software engineering and advanced applications (pp. 39–46). IEEE.
108. Jaiswal GP, Giri RN. A fuzzy inference model for reliability estimation of a component-based software system. Int J Comput Sci
Technol. 2015;3(3):177–82.
109. Mohanty R, Ravi V, Patra MR. Hybrid intelligent systems for predicting software reliability. Appl Soft Comput. 2013;13(1):189–200.
110. Tang Y, Dai Q, Yang M, Chen L, Du Y. Software defect prediction ensemble learning algorithm based on 2-step sparrow optimizing
extreme learning machine. Cluster Comput. 2024;27:1–30.
111. Zheng J. Predicting software reliability with neural network ensembles. Expert Syst Appl. 2009;36(2):2116–22.
112. Ardabili S, Mosavi A, Varkonyi-Koczy AR. 2019. Advances in machine learning modeling reviewing hybrid and ensemble methods.
113. Mosquera M, Hurtado R. 2024. Software defect prediction: a machine learning approach with voting ensemble. In International
Congress on Information and Communication Technology (pp. 585–595). Singapore: Springer Nature Singapore.
114. Kiran NR, Ravi V. Software reliability prediction by soft computing techniques. J Syst Softw. 2008;81(4):576–83.
115. Bal PR, Jena N, Mohapatra DP. 2017. Software reliability prediction based on ensemble models. In Proceeding of International Con-
ference on Intelligent Communication, Control and Devices: ICICCD 2016 (pp. 895–902). Springer Singapore.
116. Sahar O, Latif MA, Imran M. Machine learning techniques for the evaluation of efficiency of the software reliability growth models.
Gomal Univ J Res. 2017;33(1):1.
117. Li K, Zhao K, Liu W. 2013. Neural network ensemble based on K-means clustering individual selection and application for software
reliability prediction. In 2013 Fourth World Congress on Software Engineering (pp. 131–135). IEEE.
118. Ali M, Mazhar T, Arif Y, Al-Otaibi S, Ghadi YY, Shahzad T, Hamam H. 2024. Software defect prediction using an intelligent ensemble-
based model. IEEE Access.
119. Aljahdali SH, El-Telbany ME. Genetic algorithms for optimizing ensemble of models in software reliability prediction. ICGST-AIML J.
2008;8(1):5–13.
120. Elish MO, Helmy T, Hussain MI. 2013. Empirical study of homogeneous and heterogeneous ensemble models for software develop-
ment effort estimation. Math Problems Eng, 2013.
121. Behera AK, Panda, M. (2020). Software reliability prediction with ensemble method and virtual data point incorporation. In Biologi-
cally Inspired Techniques in Many-Criteria Decision Making: International Conference on Biologically Inspired Techniques in Many-
Criteria Decision Making (BITMDM-2019) (pp. 69–77). Springer International Publishing.

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z
Review

122. Gupta M, Rajnish K, Bhattacharya V. 2023. Effectiveness of ensemble classifier over state-of-art machine learning classifiers for pre-
dicting software faults in software modules. In Machine Learning, Image Processing, Network Security and Data Sciences: Select
Proceedings of 3rd International Conference on MIND 2021 (pp. 77–88). Singapore: Springer Nature Singapore.
123. Balaram A, Vasundra S. Software fault detection using multi-distinguished-features sampling with ensemble random forest classifier.
Int J Intell Eng Syst. 2022;15(5):6.
124. Borandag E. Software fault prediction using an RNN-based deep learning approach and ensemble machine learning techniques.
Appl Sci. 2023;13(3):1639.
125. Aftab S, Abbas S, Ghazal TM, Ahmad M, Hamadi HA, Yeun CY, Khan MA. A cloud-based software defect prediction system using data
and decision-level machine learning fusion. Mathematics. 2023;11(3):632.
126. Abdu A, Zhai Z, Abdo HA, Algabri R. Software defect prediction based on deep representation learning of source code from contextual
syntax and semantic graph. IEEE Trans Reliab. 2024. https://​doi.​org/​10.​1109/​TR.​2024.​33549​65.
127. Wang J, Zhang C. Software reliability prediction using a deep learning model based on the RNN encoder-decoder. Reliab Eng Syst Saf.
2018;170:73–82.
128. Mohana Ramya, Y., Deepthi, K., Vamsai, A., Juhi Sai, A., Sharma, N., & Ramachandra Reddy, B. (2023). Software Fault Prediction Using Deep
Neural Networks. In Machine Vision and Augmented Intelligence: Select Proceedings of MAI 2022 (pp. 267–274). Singapore: Springer
Nature Singapore.
129. Chua CG, Goh ATC. A hybrid Bayesian back-propagation neural network approach to multivariate modeling. Int J Numer Anal Meth
Geomech. 2003;27:651–67.
130. Karunanithi N, Whitley D. 1992. Prediction of software reliability using feedforward and recurrent neural nets. In [Proceedings 1992]
IJCNN International Joint Conference on Neural Networks (Vol. 1, pp. 800–805).
131. Hu QP, Xie M, Ng SH, Levitin G. Robust recurrent neural network modeling for software fault detection and correction prediction. Reliab
Eng Syst Saf. 2007;92(3):332–40.
132. Roy P, Mahapatra GS, Rani P, Pandey SK, Dey KN. Robust feedforward and recurrent neural network-based dynamic weighted combina-
tion models for software reliability prediction. Appl Soft Comput. 2014;22:629–37.
133. Bhuyan MK, Mohapatra DP, Sethi S. 2016. Software reliability prediction using fuzzy min-max algorithm and recurrent neural network
approach. Int J Elect Comput Eng (2088–8708), 6(4).
134. Bhuyan MK, Mohapatra DP, Sethi S. Software reliability assessment using neural networks of computational intelligence based on soft-
ware failure data. Baltic J Mod Comput. 2016;4(4):1016–37.
135. Zemouri R, Patic PC. Recurrent radial basis function network for failure time series prediction. Int J Comput Inf Eng. 2010;4(12):1920–4.
136. Fan G, Diao X, Yu H, Yang K, Chen L. 2019. Software defect prediction via attention-based recurrent neural network. Sci Programm.
137. Benaddy M, El Habil B, El Meslouhi O, Krit SD. 2018. Recurrent neural network for software failure prediction. In Proceedings of the Fourth
International Conference on Engineering & MIS 2018 (pp. 1–8).
138. Tian L, Noore A. Software reliability prediction using recurrent neural network with bayesian regularization. Int J Neural Syst.
2004;14(3):165–74.
139. Behera AK, Panda M, Dehuri S. Software reliability prediction by recurrent artificial chemical link network. Int J Syst Assurance Eng Man-
age. 2021;12(6):1308–21.
140. Pemmada SK, Nayak J, Naik B. A deep intelligent framework for software risk prediction using improved firefly optimization. Neural
Comput Appl. 2023;35(26):19523–39.
141. Zhang Q, Zhang J, Feng T, Xue J, Zhu X, Zhu N, Li Z. Software defect prediction using deep q-learning network-based feature extraction.
IET Software. 2024;2024(1):3946655.
142. Diwaker C, Tomar P, Poonia RC, Singh V. Prediction of software reliability using bio-inspired soft computing techniques. J Med Syst.
2018;42(5):1–16.
143. Benala TR, Chinnababu K, Mall R, Dehuri S. 2013. A particle swarm optimized functional link artificial neural network (PSO-FLANN) in
software cost estimation. In Proceedings of the International Conference on Frontiers of Intelligent Computing: Theory and Applications
(FICTA) (pp. 59–66). Springer, Berlin, Heidelberg.
144. Saif SM, Kirmani MM, Wahid A. 2014. Performance analysis of different software reliability prediction methods. In International Confer-
ence on Computing and Communication Technologies (pp. 1–5). IEEE.
145. Pati J, Shukla KK. 2015. A hybrid technique for software reliability prediction. In ISEC’15, February 18–20.
146. Malhotra R, Negi A. Reliability modeling using particle swarm optimization. Int J Syst Assurance Eng Manage. 2013;4(3):275–83.
147. Bisi M, Goyal NK. Prediction of software inter-failure times using artificial neural network and particle swarm optimisation models. Int J
Softw Eng Technol Appl. 2015;1(2–4):222–44.
148. Al Gargoor RG, Saleem NN. Software reliability prediction using artificial techniques. Int J Comput Sci Issues. 2013;10(4):274.
149. Yang L, Li Z, Wang D, Miao H, Wang Z. Software defects prediction based on hybrid particle swarm optimization and sparrow search
algorithm. Ieee Access. 2021;9:60865–79.
150. Qin, L. N. (2011). Software reliability prediction model based on PSO and SVM. In 2011 International Conference on Consumer Electron-
ics, Communications and Networks (CECNet) (pp. 5236–5239). IEEE. https://​doi.​org/​10.​1109/​CECNET.​2011.​57682​85.
151. Zhen L, Liu Y, Dongsheng W, Wei Z. Parameter estimation of software reliability model and prediction based on hybrid wolf pack algo-
rithm and particle swarm optimization. IEEE Access. 2020;8:29354–69.
152. Behera AK, Panda M, Dehuri S. A recurrent ANFIS tuned by modified differential evolution for efficient prediction of software reliability.
Evolut Intell. 2024;17:1–14.
153. Juneja K. A fuzzy-filtered neuro-fuzzy framework for software fault prediction for inter-version and inter-project evaluation. Appl Soft
Comput. 2019;77:696–713.
154. Yuan D, Zhang C. 2011. Evaluation strategy for software reliability based on ANFIS. In 2011 International Conference on Electronics,
Communications and Control (ICECC) (pp. 3738–3741). IEEE.
155. Zhao L, Zhang JP, Yang J, Chu Y. 2010. Software reliability growth model based on fuzzy wavelet neural network. In 2010 2nd international
conference on future computer and communication (Vol. 1, pp. V1–664). IEEE.

Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Review
Discover Computing (2025) 28:90 | https://doi.org/10.1007/s10791-025-09597-z

156. Shanmugam L, Florence L. Enhancement and comparison of ant colony optimization for software reliability models. J Comput Sci.
2013;9(9):1232.
157. Babaie E. A novel method for software reliability assessment via neuro-fuzzy system. Int J Reliab Risk Safety Theory Appl. 2022;5(1):43–8.
158. Benaddy M, Wakrim M. Simulated annealing neural network for software failure prediction. Int J Softw Eng Appl. 2012;6:4.
159. Roy P, Mahapatra GS, Dey KN. Neuro-genetic approach on logistic model-based software reliability prediction. Expert Syst Appl.
2015;42(10):4709–18.
160. Khurshid S, Iqbal J, Malik IA, Yousuf B. Modelling of NHPP based software reliability growth model from the perspective of testing cover-
age, error propagation and fault withdrawal efficiency. Int J Reliab Quality Safety Eng. 2022. https://​doi.​org/​10.​1142/​S0218​53932​25001​
39.
161. Mallikharjuna RK, Kodali A. An efficient method for enhancing reliability and selection of software reliability growth model through
optimization techniques. JSW. 2017;12(1):1–8.
162. Raamesh L, Jothi S, Radhika S. Enhancing software reliability and fault detection using hybrid brainstorm optimization-based LSTM
model. IETE J Res. 2022;3:1–15.
163. Tyagi K, Sharma A. An adaptive neuro-fuzzy model for estimating the reliability of component-based software systems. Appl Comput
Inf. 2014;10(1–2):38–51.
164. Zhang J, Bai J, Zhang Z, Feng W. Operation state assessment of wind power system based on PSO+ AHP—FCE. Front Energy Res. 2022;10:
916852.
165. Kim T, Lee K, Baik J. An effective approach to estimating the parameters of software reliability growth models using a real-valued genetic
algorithm. J Syst Softw. 2015;102:134–44.
166. Amin A, Grunske L, Colman A. An approach to software reliability prediction based on time series modeling. J Syst Softw.
2013;86(7):1923–32.
167. Umoeka IJ, Akwukwuma VN. 2023. Interval type-2 fuzzy logic system for early software reliability prediction. J Fuzzy Extension Appl.
168. Bhuyan MK, Mohapatra DP, Sethi S. 2016. Prediction strategy for software reliability based on recurrent neural network. In computational
intelligence in data mining—2 (pp. 295–303). Springer, New Delhi.
169. Gupta M, Rajnish K, Bhattacharjee V. Software fault prediction with imbalanced datasets using SMOTE-Tomek sampling technique and
Genetic Algorithm models. Multimed Tool Appl. 2023;83:1–22.
170. Meher JP, Biswas S, Mall R. Deep learning-based software bug classification. Inf Softw Technol. 2024;166: 107350.
171. Samal U, Kumar A. Enhancing software reliability forecasting through a hybrid ARIMA-ANN model. Arabian J Sci Eng. 2023;4:1–14.
172. Azzeh M, Nassif AB, Talib MA, Iqba H. Software defect prediction using non-dominated sorting genetic algorithm and k-nearest neighbour
classifier. e-Inf Softw Eng J. 2024;18(1):240103.
173. Costa EO, Pozo ATR, Vergilio SR. A genetic programming approach for software reliability modeling. IEEE Trans Reliab. 2010;59(1):222–30.
174. Liu Z, Yang S, Yang M, Kang R. Software belief reliability growth model based on uncertain differential equation. IEEE Trans Reliab.
2022;71(2):775–87.
175. Li S, Dohi T, Okamura H. Nonhomogeneous Markov process modeling for software reliability assessment. IEEE Trans Reliab.
2023;72(4):1540–55.
176. Zama W, Xiao X. 2023. Software reliability prediction via neural network. In 2023 IEEE 34th International Symposium on Software Reli-
ability Engineering Workshops (ISSREW) (pp. 103–107). IEEE.
177. Dhavakumar P, Gopalan NP. An efficient parameter optimization of software reliability growth model by using chaotic grey wolf opti-
mization algorithm. J Ambient Intell Humaniz Comput. 2021;12:3177–88.
178. Chatterjee S, Chaudhuri B, Bhar C. Optimal release time determination via fuzzy goal programming approach for SDE-based software
reliability growth model. Soft Comput. 2021;25(5):3545–64.
179. Wang X, Wu J, Liu C, Wang S, Niu W. A hybrid model based on singular spectrum analysis and support vector machines regression for
failure time series prediction. Qual Reliab Eng Int. 2016;32(8):2717–38.
180. Sobhana M, Preethi GS, Sri GH, Sujitha KB. 2022. Improved reliability prediction in engineering systems based on artificial neural network.
In 2022 International Mobile and Embedded Technology Conference (MECON) (pp. 455–460). IEEE.
181. Han MF, Lin CT, Chang JY. Differential evolution with local information for neuro-fuzzy systems optimisation. Knowl-Based Syst.
2013;44:78–89.
182. Liu L, Jiang Z. 2016. Research on software reliability evaluation technology based on BP neural network. In 2016 IEEE/ACIS 15th Inter-
national Conference on Computer and Information Science (ICIS) (pp. 1–4). IEEE.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Vol:.(1234567890)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

[email protected]