On the use of deep learning in software defect prediction

Journal of Systems and Software, Год журнала: 2022, Номер 195, С. 111537 - 111537

Опубликована: Окт. 12, 2022

Automated software defect prediction (SDP) methods are increasingly applied, often with the use of machine learning (ML) techniques. Yet, existing ML-based approaches require manually extracted features, which cumbersome, time consuming and hardly capture semantic information reported in bug reporting tools. Deep (DL) techniques provide practitioners opportunities to automatically extract learn from more complex high-dimensional data. The purpose this study is systematically identify, analyze, summarize, synthesize current state utilization DL algorithms for SDP literature. We selected a pool 102 peer-reviewed studies then conducted quantitative qualitative analysis using data these studies. Main highlights include: (1) most applied supervised DL; (2) two third used metrics as an input algorithms; (3) Convolutional Neural Network frequently algorithm. Based on our findings, we propose develop comprehensive that needed features; diverse artifacts other than source code; adopt augmentation tackle class imbalance problem; (4) publish replication packages.

Язык: Английский

---

Predicting test failures induced by software defects: A lightweight alternative to software defect prediction and its industrial application

Journal of Systems and Software, Год журнала: 2025, Номер 223, С. 112360 - 112360

Опубликована: Фев. 3, 2025

Язык: Английский

---

Code Smell Detection Using Ensemble Machine Learning Algorithms

Applied Sciences, Год журнала: 2022, Номер 12(20), С. 10321 - 10321

Опубликована: Окт. 13, 2022

Code smells are the result of not following software engineering principles during development, especially in design and coding phase. It leads to low maintainability. To evaluate quality its maintainability, code smell detection can be helpful. Many machine learning algorithms being used detect smells. In this study, we applied five ensemble two deep Four datasets were analyzed: Data class, God Feature-envy, Long-method datasets. previous works, stacking dataset results found acceptable, but there is scope improvement. A class balancing technique (SMOTE) was handle imbalance problem The Chi-square feature extraction select more relevant features each dataset. All obtained highest accuracy—100% for with different selected sets metrics, poorest accuracy, 91.45%, achieved by Max voting method Feature-envy twelve metrics.

Язык: Английский

---

A systematic review on food recommender systems

Expert Systems with Applications, Год журнала: 2023, Номер 238, С. 122166 - 122166

Опубликована: Окт. 16, 2023

The Internet has revolutionized the way information is retrieved, and increase in number of users resulted a surge volume heterogeneity available data. Recommender systems have become popular tools to help retrieve relevant quickly. Food Systems (FRS), particular, proven useful overcoming overload present food domain. However, recommendation complex domain with specific characteristics causing many challenges. Additionally, very few systematic literature reviews been conducted on FRS. This paper presents review that summarizes current state-of-the-art Our examines different methods algorithms used for recommendation, data how it processed, evaluation methods. It also advantages disadvantages To achieve this, total 67 high-quality studies were selected from pool 2,738 using strict quality criteria. provides valuable research field, helping researchers select strategy develop can improve efficiency development, thus closing gap between development FRS other recommender systems.

Язык: Английский

---

Detecting code smells using industry-relevant data

Information and Software Technology, Год журнала: 2022, Номер 155, С. 107112 - 107112

Опубликована: Ноя. 21, 2022

Язык: Английский

---

A critical comparison on six static analysis tools: Detection, agreement, and precision

Journal of Systems and Software, Год журнала: 2022, Номер 198, С. 111575 - 111575

Опубликована: Ноя. 30, 2022

Developers use Static Analysis Tools (SATs) to control for potential quality issues in source code, including defects and technical debt. Tool vendors have devised quite a number of tools, which makes it harder practitioners select the most suitable one their needs. To better support developers, researchers been conducting several studies on SATs favor understanding actual capabilities. Despite work done so far, there is still lack knowledge regarding (1) what agreement, (2) precision recommendations. We aim at bridging this gap by proposing large-scale comparison six popular Java projects: Better Code Hub, CheckStyle, Coverity Scan, FindBugs, PMD, SonarQube. analyze 47 projects applying 6 SATs. assess we compared them manually analyzing – line class-level — whether they identify same issues. Finally, evaluate tools against manually-defined ground truth. The key results show little no agreement among low degree precision. Our study provides first overview different as well an extensive analysis that can be used researchers, practitioners, tool map current capabilities envision possible improvements.

Язык: Английский

---

FedCSD: A Federated Learning Based Approach for Code-Smell Detection

IEEE Access, Год журнала: 2024, Номер 12, С. 44888 - 44904

Опубликована: Янв. 1, 2024

Software quality is critical, as low quality, or "Code smell," increases technical debt and maintenance costs. There a timely need for collaborative model that detects manages code smells by learning from diverse distributed data sources while respecting privacy providing scalable solution continuously integrating new patterns practices in management. However, the current literature still missing such capabilities. This paper addresses previous challenges proposing Federated Learning Code Smell Detection (FedCSD) approach, specifically targeting "God Class," to enable organizations train ML models safeguarding collaboratively. We conduct experiments using manually validated datasets detect analyze smell scenarios validate our approach. Experiment 1, centralized training experiment, revealed varying accuracies across datasets, with dataset two achieving lowest accuracy (92.30%) one three highest (98.90% 99.5%, respectively). 2, focusing on cross-evaluation, showed significant drop (lowest: 63.80%) when fewer were present dataset, reflecting debt. 3 involved splitting 10 companies, resulting global of 98.34%, comparable model's accuracy. The application federated techniques demonstrates promising performance improvements code-smell detection, benefiting both software developers researchers.

Язык: Английский

---

Automatic detection of code smells using metrics and CodeT5 embeddings: a case study in C#

Neural Computing and Applications, Год журнала: 2024, Номер 36(16), С. 9203 - 9220

Опубликована: Фев. 24, 2024

Язык: Английский

---

Dlap: A Deep Learning Augmented Large Language Model Prompting Framework for Software Vulnerability Detection

Опубликована: Янв. 1, 2024

Software vulnerability detection is supported by automated static analysis tools, which have recently been reinforced deep learning (DL) models. However, despite the superior performance of DL-based approaches over rule-based ones in research, applying DL to software practice remains a challenge due complex structure source code, black-box nature DL, and domain knowledge required understand validate results for addressing tasks after detection. Conventional models are trained specific projects and, hence, excel identifying vulnerabilities these but not others. These with poor would impact downstream such as location repair. do provide explanations developers comprehend results. In contrast, Large Language Models (LLMs) made lots progress issues leveraging prompting techniques. Unfortunately, their unsatisfactory. This paper contributes \textbf{\DLAP}, \underline{\textbf{D}}eep \underline{\textbf{L}}earning \underline{\textbf{A}}ugmented LLMs \underline{\textbf{P}}rompting framework that combines both achieve exceptional performance. Experimental evaluation confirm \DLAP outperforms state-of-the-art frameworks fine-turning on multiple metrics.

Язык: Английский

---

DLAP: A deep learning augmented Large Language Model prompting framework for software vulnerability detection

Journal of Systems and Software, Год журнала: 2024, Номер 219, С. 112234 - 112234

Опубликована: Окт. 18, 2024

Язык: Английский

---