Software defect prediction using a bidirectional LSTM network combined with oversampling techniques

Cluster Computing, Journal Year: 2023, Volume and Issue: 27(3), P. 3615 - 3638

Published: Oct. 28, 2023

Abstract Software defects are a critical issue in software development that can lead to system failures and cause significant financial losses. Predicting is vital aspect of ensuring quality. This significantly impact both saving time reducing the overall cost testing. During defect prediction (SDP) process, automated tools attempt predict source codes based on metrics. Several SDP models have been proposed identify prevent before they occur. In recent years, recurrent neural network (RNN) techniques gained attention for their ability handle sequential data learn complex patterns. Still, these not always suitable predicting due problem imbalanced data. To deal with this problem, study aims combine bidirectional long short-term memory (Bi-LSTM) oversampling techniques. establish effectiveness efficiency model, experiments conducted benchmark datasets obtained from PROMISE repository. The experimental results compared evaluated terms accuracy, precision, recall, f-measure, Matthew’s correlation coefficient (MCC), area under ROC curve (AUC), precision-recall (AUCPR) mean square error (MSE). average accuracy model original balanced (using random SMOTE) was 88%, 94%, And 92%, respectively. showed Bi-LSTM improves by 6 4% datasets. F-measure were 51%, 43 41% demonstrated combining positively affects performance class distributions.

Language: Английский

---

Cervical Cancer Diagnosis Using Stacked Ensemble Model and Optimized Feature Selection: An Explainable Artificial Intelligence Approach

Computers, Journal Year: 2023, Volume and Issue: 12(10), P. 200 - 200

Published: Oct. 7, 2023

Cervical cancer affects more than half a million women worldwide each year and causes over 300,000 deaths. The main goals of this paper are to study the effect applying feature selection methods with stacking models for prediction cervical cancer, propose ensemble learning that combines different meta-learners predict explore black-box model best-optimized features using explainable artificial intelligence (XAI). A dataset from machine repository (UCI) is highly imbalanced contains missing values used. Therefore, SMOTE-Tomek was used combine under-sampling over-sampling handle data, pre-processing steps implemented hold values. Bayesian optimization optimizes selects best architecture. Chi-square scores, recursive removal, tree-based three techniques applied For determining factors most crucial predicting extended multiple levels: Level 1 (multiple base learners) 2 (meta-learner). At 1, (training testing stacking) employed combining output multi-base models, while training train meta-learner at level 2. Testing evaluate models. results showed based on selected elimination (RFE), has higher accuracy, precision, recall, f1-score, AUC. Furthermore, To assure efficiency, efficacy, reliability produced model, local global explanations provided.

Language: Английский

---

Ensemble-Based Machine Learning Algorithms Combined with Near Miss Method for Software Bug Prediction

˜The œInternational journal of networked and distributed computing, Journal Year: 2025, Volume and Issue: 13(1)

Published: Jan. 13, 2025

Abstract Software bug prediction (SBP) involves identifying or categorizing software modules likely to contain defects, utilizing underlying system properties such as metrics. SBP plays a crucial role in enhancing project quality and mitigating maintenance risks. Numerous machine learning (ML) algorithms have been developed predict bugs. Class imbalance poses significant challenge for these algorithms, significantly impeding their effectiveness resulting imbalanced false-positive false-negative outcomes. However, limited research has conducted specifically tackle the issue of class context SBP. This study investigates performance homogeneous ensemble: Bagging, boosting, voting classifiers (VC) methods combined with under-sampling address problem improve accuracy Two ensembles are classified bagging ensembles: decision tree (DT) random forest (RF); two boosting AdaBoost (AB) gradient (GB), while DT, RF, K-Nearest Neighbours (K-NN), support vector (SVM) considered VC. To establish proposed models, experiments were on available benchmark datasets, which comprise five public datasets based both file-level We compared evaluated models according several measures, namely accuracy, precision, recall, f-measure, Matthew’s correlation coefficient (MCC), area under receiver operating characteristic curve (AUROC). The experimental findings demonstrated that exhibit superior efficiency predicting bugs balanced original an improvement up 11% class-level metrics 10% results indicate use data sampling techniques had positive impact presented models. our method existing standard measures. comparison outcomes revealed superiority over prevailing state-of-the-art across most datasets.

Language: Английский

---

Adaptive Ensemble Learning Model-Based Binary White Shark Optimizer for Software Defect Classification

International Journal of Computational Intelligence Systems, Journal Year: 2025, Volume and Issue: 18(1)

Published: Jan. 23, 2025

Language: Английский

---

Capsule feature selector for software defect prediction

The Journal of Supercomputing, Journal Year: 2025, Volume and Issue: 81(3)

Published: Feb. 13, 2025

Language: Английский

---

Software defect prediction based on residual/shuffle network optimized by upgraded fish migration optimization algorithm

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 28, 2025

The study introduces a new method for predicting software defects based on Residual/Shuffle (RS) Networks and an enhanced version of Fish Migration Optimization (UFMO). overall contribution is to improve the accuracy, reduce manual effort needed. originality this work rests in synergic use deep learning metaheuristics train code extraction semantic structural properties. model tested variety open-source projects, yielding average accuracy 93% surpassing performance state-of-the-art models. results indicate increase precision (78–98%), recall (71–98%), F-measure (72–96%), Area Under Curve (AUC) (78–99%). proposed simple efficient proves be effective identifying potential defects, consequently decreasing chance missing these improving quality as opposed existing approaches. However, analysis limited projects warrants further evaluation proprietary software. enables robust tool developers. This approach can revolutionize development practices order artificial intelligence solve difficult issues presented offers high cost, which user satisfaction, enhance being developed.

Language: Английский

---

Just-in-time software defect location for mobile applications based on pre-trained programming encoder model and hierarchical attention network

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 148, P. 110381 - 110381

Published: Feb. 27, 2025

Language: Английский

---

Machine Learning Approaches in Software Vulnerability Detection: A Systematic Review and Analysis of Contemporary Methods

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: March 21, 2025

This systematic review examines the application of machine learning (ML) techniques in software vulnerability detection, focusing on their effectiveness identifying, classifying, and mitigating security risks within source code. The synthesizes findings from 83 studies published between 2019 2024, encompassing static, dynamic, hybrid detection methods. Key objectives include categorizing ML ap-plications for specific tasks, evaluating methodologies, com-piling relevant datasets tools, identifying challenges opportunities field. Findings reveal a predominant reliance static analysis techniques, supported by advanced models such as Graph Neural Networks (GNNs) transformer-based Natural Language Processing (NLP) like CodeBERT. While deep approaches dominate due to ability process large-scale complex patterns, traditional methods remain significant contexts requiring interpretability smaller datasets. Analysis highlights focus C/C++ programming family, with substantial dataset diversity, scalability, class imbalances. Opportunities improvement integration multilingual datasets, static-dynamic methods, architectures enhance accuracy reduce computational overhead. identifies need explainable AI, real-world validation, user friendly tools bridge gap academic research industrial application. By addressing these challenges, future advancements ML-based can contribute development scalable, interpretable, effective solutions modern security.

Language: Английский

---

Integrating temporal convolutional networks with metaheuristic optimization for accurate software defect prediction

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(5), P. e0319562 - e0319562

Published: May 12, 2025

The increasing importance of deep learning in software development has greatly improved quality by enabling the efficient identification defects, a persistent challenge throughout lifecycle. This study seeks to determine most effective model for detecting defects projects. It introduces an intelligent approach that combines Temporal Convolutional Networks (TCN) with Antlion Optimization (ALO). TCN is employed defect detection, while ALO optimizes network’s weights. Two models are proposed address research problem: (a) basic without parameter optimization and (b) hybrid integrating ALO. findings demonstrate significantly outperforms multiple performance metrics, including area under curve, sensitivity, specificity, accuracy, error rate. Moreover, surpasses state-of-the-art methods, such as Neural Networks, Gated Recurrent Units, Bidirectional Long Short-Term Memory, accuracy improvements 21.8%, 19.6%, 31.3%, respectively. Additionally, achieves 13.6% higher curve across all datasets compared Deep Forest method. These results confirm effectiveness accurately diverse

Language: Английский

---

An Uncertainty-Incorporated Active Data Diffusion Learning Framework for Few-Shot Equipment RUL Prediction

Reliability Engineering & System Safety, Journal Year: 2024, Volume and Issue: 254, P. 110632 - 110632

Published: Nov. 13, 2024

Language: Английский

---