软件缺陷预测旨在识别软件开发过程中的高风险缺陷模块，以实现资源的高效分配。近年来，基于图神经网络的缺陷预测方法因其能够更全面地捕捉模块间的交互关系，受到了广泛关注和研究。随着软件系统日益复杂以及开发团队规模日益扩大，开发者的行为特征和协作模式对缺陷的产生影响愈加显著。然而，现有相关方法在构图方法上趋于同质化，通常仅依赖单一的代码视角构建同构图，忽视了开发者因素在软件开发中的作用。此外，目前缺少采用异构图形式进行缺陷预测的方法。如何突破当前构图方法的限制，在图结构上实现有效创新，成为当前研究的新挑战。

　　本文提出了两种基于图神经网络的软件缺陷预测方法，分别结合开发者与代码的双重视角，构建了两种新颖的双视角同构图和异构图，并将其应用于图神经网络中，在实际应用背景下进行了验证。本文的主要贡献如下：

（1）针对当前软件依赖图视角单一的问题，本文提出了基于双视角同构图的图神经网络缺陷预测方法（DeDuVGN）。该方法提出了一种双视角软件依赖图构建策略，通过整合代码依赖和开发者依赖关系，旨在更加全面地刻画软件系统中的多维关系。此外，DeDuVGN结合了少数类过采样技术与双向门控图神经网络，提升了模型对少数类缺陷的识别能力。实验结果表明，DeDuVGN在多个开源软件项目的多项评价指标上显著优于现有方法，F1得分较当前最先进方法提升了10.7%。

  （2）针对当前缺乏利用异构图进行缺陷预测的现状，本文在同构图结构上进一步扩展，提出了基于双视角异构图的图神经网络缺陷预测方法（DeDVHeGN）。该模型通过在异构图中引入代码模块和开发者两种节点类型，并定义多种依赖边类型，以更全面地反映软件系统中的复杂依赖关系。本方法还总结出一套多维度的开发者特征提取方法，并利用图节点采样策略改进异构图神经网络，有效缓解了缺陷类别不平衡问题。实验结果表明，DeDVHeGN在多个标准数据集上优于其他先进模型，F1得分较当前最先进方法提升了13.3%。同时，本文还比较了DeDuVGN与DeDVHeGN两种方法的优劣，讨论了各自适用的场景。

  （3）设计并实现了基于上述两种方法的软件缺陷预测系统。该系统提供了一个可靠的应用工具，将上述方法封装为功能模块，并通过可视化交互界面呈现，旨在降低缺陷预测相关人员的学习门槛，使缺陷预测过程更加直观、高效和便捷。

Software defect prediction aims to identify high-risk defect modules during the software development process to enable efficient resource allocation. In recent years, defect prediction methods based on Graph Neural Networks (GNNs) have attracted widespread attention and research due to their ability to more comprehensively capture the interactions between modules, thus improving prediction performance. With the increasing complexity of software systems and the expanding scale of development teams, the influence of developer behaviors and collaboration patterns on defect generation has become more significant. However, existing methods are becoming more homogeneous in their graph construction approaches, typically relying solely on code-based perspectives to build homogeneous graphs, neglecting the role of developer factors in software development. Furthermore, there is a lack of methods using heterogeneous graphs for defect prediction. Overcoming the limitations of current graph construction methods and innovating in graph structure design has become a new challenge in the field.

The thesis proposes two Graph Neural Network-based software defect prediction methods, each combining a dual-view approach that incorporates both developer and code features. Two novel dual-view homogeneous and heterogeneous graphs are constructed and applied to GNNs, with practical verification conducted in real-world application scenarios. The main contributions of the thesis are as follows:

(1) To address the issue of a single perspective in current software dependency graphs, the thesis proposes a graph neural network defect prediction method based on dual-view homogeneous graphs (DeDuVGN). This method introduces a dual-view software dependency graph construction approach by integrating code dependencies and developer dependencies, aiming to provide a more comprehensive representation of multi-dimensional relationships in a software system. In addition, DeDuVGN combines minority class oversampling techniques with bidirectional gated graph neural networks to effectively address the data imbalance problem, improving the model's ability to identify minority defects. Experimental results show that DeDuVGN significantly outperforms existing methods on multiple evaluation metrics across several open-source software projects, with the F1 score improving by 10.7% compared to the current state-of-the-art methods.

(2) In response to the lack of defect prediction methods using heterogeneous graphs, the thesis further extends the homogeneous graph structure and proposes a graph neural network defect prediction method based on dual-view heterogeneous graphs (DeDVHeGN). This model introduces two node types—code modules and developers—into the heterogeneous graph and defines multiple types of dependency edges to more comprehensively reflect the complex dependencies in a software system. The method also develops a multi-dimensional developer feature extraction approach and improves heterogeneous graph neural networks using a graph node sampling strategy to effectively mitigate the data imbalance issue. Experimental results show that DeDVHeGN outperforms other advanced models on multiple benchmark datasets, with the F1 score improving by 13.3% compared to the current state-of-the-art methods. Additionally, the thesis compares the strengths and weaknesses of DeDuVGN and DeDVHeGN, discussing their respective applicable scenarios.

(3) A software defect prediction system based on the above two methods is designed and implemented. The system provides a reliable application tool by encapsulating the aforementioned methods into functional modules and presenting them through a visual interactive interface. The goal is to lower the learning threshold for defect prediction personnel and make the defect prediction process more intuitive, efficient, and convenient.

[1] Tian X, Chang J, Zhang C, et al. 开源软件缺陷预测方法综述 [J]. Jisuanji Yanjiu yu Fazhan/Computer Research and Development, 2023, 60(7):1467–1488.

[2] 仇正霞, 杨剑锋, 胡文生, et al. 基于光滑样条回归的软件可靠性模型 [J]. Modeling and

Simulation, 2023, 12:5156.

[3] Mugu S R, Zhang B, Kolla H, et al. Lessons from the CrowdStrike Incident: Assessing End-

point Security Vulnerabilities and Implications[C]. Proceedings of 2024 Cyber Awareness and

Research Symposium (CARS). IEEE, 2024. 1–10.

[4] Jones C. Applied software measurement: assuring productivity and quality[M]. McGraw-Hill,Inc., 1991.

[5] Fenton N E, Neil M. A critique of software defect prediction models[J]. IEEE Transactions onsoftware engineering, 1999, 25(5):675–689.

[6] Thota M K, Shajin F H, Rajesh P, et al. Survey on software defect prediction techniques[J].

International Journal of Applied Science and Engineering, 2020, 17(4):331–344.

[7] 邓枭, 叶蔚, 谢睿, et al. 基于深度学习的源代码缺陷检测研究综述 [J]. 软件学报, 2023,

34(2):625–654.

[8] Çalıklı G, Bener A B. Influence of confirmation biases of developers on software quality: an

empirical study[J]. Software Quality Journal, 2013, 21:377–416.

[9] Datta S. How does developer interaction relate to software quality? An examination of product development data[J]. Empirical Software Engineering, 2018, 23(3):1153–1187.

[10] Hsu H C, Lin T L, Wu B J, et al. FincGAN: A Gan Framework Performance analysis of machine

learning techniques on software defect prediction using NASA datasets Imbalanced Node Classification on Heterogeneous Graph Neural Network[C]. Proceedings of ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024.

5750–5754.

[11] Elish K O, Elish M O. Predicting defect-prone software modules using support vector ma-

chines[J]. Journal of Systems and Software, 2008, 81(5):649–660.

[12] Perreault L, Berardinelli S, Izurieta C, et al. Using classifiers for software defect detection[C].

Proceedings of 26th International conference on software engineering and data engineering, 2017. 2–4.

[13] Gray D, Bowes D, Davey N, et al. Using the support vector machine as a classification method for software defect prediction with static code metrics[C]. Proceedings of Engineering Applications of Neural Networks: 11th International Conference, EANN 2009, London, UK, August 27-29, 2009. Proceedings 11. Springer, 2009. 223–234.

[14] Lessmann S, Baesens B, Mues C, et al. Benchmarking classification models for software defect prediction: A proposed framework and novel findings[J]. IEEE transactions on software

engineering, 2008, 34(4):485–496.

[15] Pushphavathi T, Suma V, Ramaswamy V. A novel method for software defect prediction: hybrid of FCM and random forest[C]. Proceedings of 2014 International Conference on Electronics and Communication Systems (ICECS). IEEE, 2014. 1–5.

[16] Wang T, Li W h. Naive bayes software defect prediction model[C]. Proceedings of 2010 International conference on computational intelligence and software engineering. Ieee, 2010. 1–4.

[17] Chen T, Guestrin C. Xgboost: A scalable tree boosting system[C]. Proceedings of Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 2016. 785–794.

[18] Wahono R S. A systematic literature review of software defect prediction[J]. Journal of software engineering, 2015, 1(1):1–16.

[19] Guyon I, Elisseeff A. An introduction to variable and feature selection[J]. Journal of machine learning research, 2003, 3(Mar):1157–1182.

[20] Gao Y, Yang C. Software defect prediction based on adaboost algorithm under imbalance distribution[C]. Proceedings of 2016 4th International Conference on Sensors, Mechatronics and Automation (ICSMA 2016). Atlantis Press, 2016. 739–746.

[21] Cao Q, Sun Q, Cao Q, et al. Software defect prediction via transfer learning based neural network[C]. Proceedings of 2015 First international conference on reliability systems engineering (ICRSE). IEEE, 2015. 1–10.

[22] Goodfellow I. Deep learning, 2016.

[23] Apicella A, Donnarumma F, Isgrò F, et al. A survey on modern trainable activation functions[J]. Neural Networks, 2021, 138:14–32.

[24] Golovko V, Kroshchanka A, Rubanau U, et al. A Learning Technique for Deep Belief Neural

Networks[M]. Springer International Publishing, 2014: 136–146.

[25] Wang S, Liu T, Nam J, et al. Deep Semantic Feature Learning for Software Defect Prediction[J]. IEEE Transactions on Software Engineering, 2020, 46(12):1267–1293.

[26] Rao D, McMahan B. Natural language processing with PyTorch: build intelligent language applications using deep learning[M]. ” O’Reilly Media, Inc.”, 2019.

[27] Li J, He P, Zhu J, et al. Software defect prediction via convolutional neural network[C]. Proceedings of 2017 IEEE international conference on software quality, reliability and security (QRS). IEEE, 2017. 318–328.

[28] Hochreiter S, Schmidhuber J. Long Short-Term Memory[J]. Neural Computation, 1997,

9(8):1735–1780.

[29] Chung J, Gulcehre C, Cho K, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling[J]. arXiv preprint arXiv:1412.3555, 2014..

[30] Munir H S, Ren S, Mustafa M, et al. Attention based GRU-LSTM for software defect predic

tion[J]. PLOS ONE, 2021, 16(3):e0247444.

[31] Cho K. Learning phrase representations using RNN encoder-decoder for statistical machine translation[J]. arXiv preprint arXiv:1406.1078, 2014..

[32] Kumar P, Venkatesan D R. Improving Software Defect Prediction using Generative Adversarial Networks[J]. International journal of Science and Engineering Applications, 2020, 9(9):117–120.

[33] Zhao L, Shang Z, Zhao L, et al. Software defect prediction via cost-sensitive Siamese parallel fully-connected neural networks[J]. Neurocomputing, 2019, 352:64–74.

[34] Yu H, Sun X, Zhou Z, et al. A novel software defect prediction method based on hierarchical neu- ral network[C]. Proceedings of 2021 IEEE 45th Annual Computers, Software, and Applications Conference (COMPSAC). IEEE, 2021. 366–375.

[35] Duan X, Wu J, Ji S, et al. VulSniper: Focus Your Attention to Shoot Fine-Grained Vulnerabilities.[C]. Proceedings of IJCAI, 2019. 4665–4671.

[36] Zhou Y, Liu S, Siow J, et al. Devign: Effective vulnerability identification by learning com-

prehensive program semantics via graph neural networks[J]. Advances in neural information

processing systems, 2019, 32.

[37] Cheng X, Wang H, Hua J, et al. Deepwukong: Statically detecting software vulnerabilities using deep graph neural network[J]. ACM Transactions on Software Engineering and Methodology, 2021, 30(3):1–33.

[38] Wang H, Ye G, Tang Z, et al. Combining graph-based learning with automated data collection for code vulnerability detection[J]. IEEE Transactions on Information Forensics and Security, 2020, 16:1943–1958.

[39] Zhou C, He P, Zeng C, et al. Software defect prediction with semantic and structural information of codes based on Graph Neural Networks[J]. Information and Software Technology, 2022, 152:107057.

[40] Qiu S, Huang M, Liang Y, et al. Code Multiview Hypergraph Representation Learning for Software Defect Prediction[J]. IEEE Transactions on Reliability, 2024..

[41] Liu W, Yue Y, Chen X, et al. SeDPGK: Semi-supervised software defect prediction with graph representation learning and knowledge distillation[J]. Information and Software Technology, 2024. 107510.

[42] Mockus A, Weiss D M. Predicting risk of software changes[J]. Bell Labs Technical Journal,

2000, 5(2):169–180.

[43] Wu Y, Yang Y, Zhao Y, et al. The influence of developer quality on software fault-proneness prediction[C]. Proceedings of 2014 eighth international conference on software security and reliability (SERE). IEEE, 2014. 11–19.

[44] Karimi Z, Baraani-Dastjerdi A, Ghasem-Aghaee N, et al. Links between the personalities, styles and performance in computer programming[J]. Journal of Systems and Software, 2016, 111:228–241.

[45] Ando R, Sato S, Uchida C, et al. How does defect removal activity of developer vary with development experience?[C]. Proceedings of SEKE, 2015. 540–545.

[46] Kini S O, Tosun A. Periodic developer metrics in software defect prediction[C]. Proceedings of 2018 IEEE 18th International Working Conference on Source Code Analysis and Manipulation (SCAM). IEEE, 2018. 72–81.

[47] Hokka H, Dobslaw F, Bengtsson J. Linking developer experience to coding style in open-source repositories[C]. Proceedings of 2021 IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER). IEEE, 2021. 516–520.

[48] Piantadosi V, Scalabrino S, Serebrenik A, et al. Do attention and memory explain the performance of software developers?[J]. Empirical Software Engineering, 2023, 28(5):112.

[49] Eyolfson J, Tan L, Lam P. Do time of day and developer experience affect commit bugginess?[C]. Proceedings of Proceedings of the 8th Working Conference on Mining Software Repositories, 2011. 153–162.

[50] Qiu Y, Zhang W, Zou W, et al. An empirical study of developer quality[C]. Proceedings of 2015 IEEE International Conference on Software Quality, Reliability and Security-Companion. IEEE, 2015. 202–209.

[51] Lavallée M, Robillard P N. Why good developers write bad code: An observational case study of the impacts of organizational factors on software quality[C]. Proceedings of 2015 IEEE/ACM 37th IEEE International Conference on Software Engineering, volume 1. IEEE, 2015. 677–687.

[52] Meneely A, Williams L. Secure open source collaboration: an empirical study of linus’ law[C]. Proceedings of Proceedings of the 16th ACM conference on Computer and communications security, 2009. 453–462.

[53] Gong L, Rajbahadur G K, Hassan A E, et al. Revisiting the impact of dependency network

metrics on software defect prediction[J]. IEEE Transactions on Software Engineering, 2021,

48(12):5030–5049.

[54] Yatish S, Jiarpakdee J, Thongtanunam P, et al. Mining software defects: Should we consider affected releases?[C]. Proceedings of 2019 IEEE/ACM 41st International Conference on Software Engineering. IEEE, 2019. 654–665.

[55] He Z, Shu F, Yang Y, et al. An investigation on the feasibility of cross-project defect prediction[J]. Automated Software Engineering, 2012, 19:167–199.

[56] Rajbahadur G K, Wang S, Kamei Y, et al. Impact of discretization noise of the dependent variable on machine learning classifiers in software engineering[J]. IEEE Transactions on Software Engineering, 2019, 47(7):1414–1430.

[57] Woolson R F. Wilcoxon signed-rank test[J]. Wiley encyclopedia of clinical trials, 2007. 1–3.

[58] Macbeth G, Razumiejczyk E, Ledesma R D. Cliff’s Delta Calculator: A non-parametric effect size program for two groups of observations[J]. Universitas Psychologica, 2011, 10(2):545–555.

[59] Armstrong R A. When to use the B onferroni correction[J]. Ophthalmic and Physiological Optics, 2014, 34(5):502–508.

[60] Guan Z, Wang X, Xin W, et al. A survey on deep learning-based source code defect analysis[C]. Proceedings of 2020 5th International Conference on Computer and Communication Systems. IEEE, 2020. 167–171.

[61] Gong L, Jiang S, Jiang L. An improved transfer adaptive boosting approach for mixed-project defect prediction[J]. Journal of Software: Evolution and Process, 2019, 31(10):e2172.

[62] Prasad M, Florence L, Arya A. A study on software metrics based software defect prediction using data mining and machine learning techniques[J]. International Journal of Database Theory and Application, 2015, 8(3):179–190.

[63] Li F, Lu W, Keung J W, et al. The impact of feature selection techniques on effort-aware defect prediction: An empirical study[J]. IET Software, 2023, 17(2):168–193.

[64] Zimmermann T, Nagappan N. Predicting defects using network analysis on dependency

graphs[C]. Proceedings of Proceedings of the 30th international conference on Software en-

gineering, 2008. 531–540.

[65] Buse R P L, Weimer W R. Learning a Metric for Code Readability[J]. IEEE Transactions on

Software Engineering, 2010, 36(4):546–558.

[66] Mockus A, Weiss D M. Predicting risk of software changes[J]. Bell Labs Technical Journal,

2002, 5(2):169–180.

[67] Phan A V, Le Nguyen M, Bui L T. Convolutional neural networks over control flow graphs for software defect prediction[C]. Proceedings of 2017 IEEE 29th International Conference on Tools with Artificial Intelligence. IEEE, 2017. 45–52.

[68] Feng Q, Feng C, Hong W. Graph neural network-based vulnerability predication[C]. Proceedings of 2020 IEEE International Conference on Software Maintenance and Evolution. IEEE, 2020. 800–801.

[69] Ghaffarian S M, Shahriari H R. Neural software vulnerability analysis using rich intermediate graph representations of programs[J]. Information Sciences, 2021, 553:189–207.

[70] Mantyla M, Lassenius C. What Types of Defects Are Really Discovered in Code Reviews?[J].

IEEE Transactions on Software Engineering, 2009, 35(3):430–448.

[71] Chawla N V, Bowyer K W, Hall L O, et al. SMOTE: synthetic minority over-sampling tech-

nique[J]. Journal of artificial intelligence research, 2002, 16:321–357.

[72] Chen Y, Wu L, Zaki M J. Reinforcement learning based graph-to-sequence model for natural question generation[J]. arXiv preprint arXiv:1908.04942, 2019..

[73] Liu S, Xie X, Siow J, et al. Graphsearchnet: Enhancing gnns via capturing global dependencies for semantic code search[J]. IEEE Transactions on Software Engineering, 2023..

[74] Wang S, Liu T, Nam J, et al. Deep semantic feature learning for software defect prediction[J]. IEEE Transactions on Software Engineering, 2018, 46(12):1267–1293.

[75] Arar Ö F, Ayan K. A feature dependent Naive Bayes approach and its application to the software defect prediction problem[J]. Applied Soft Computing, 2017, 59:197–209.

[76] Yi J, Kim B, Chang B. Embedding Normalization: Significance Preserving Feature Normal-

ization for Click-Through Rate Prediction[C]. Proceedings of 2021 International Conference on Data Mining Workshops. IEEE, 2021. 75–84.

[77] Malhotra R. A systematic review of machine learning techniques for software fault prediction[J]. Applied Soft Computing, 2015, 27:504–518.

[78] Turhan B, Mısırlı A T, Bener A. Empirical evaluation of the effects of mixed project data on learning defect predictors[J]. Information and Software Technology, 2013, 55(6):1101–1118.

[79] Aleem S, Capretz L F, Ahmed F. Benchmarking machine learning technologies for software defect detection[J]. arXiv preprint arXiv:1506.07563, 2015..

[80] Miao L, Liu M, Zhang D. Cost-sensitive feature selection with application in software defect prediction[C]. Proceedings of Proceedings of the 21st international conference on pattern recognition. IEEE, 2012. 967–970.

[81] Rajbahadur G K, Wang S, Kamei Y, et al. The impact of using regression models to build defect classifiers[C]. Proceedings of 2017 IEEE/ACM 14th International Conference on Mining

Software Repositories. IEEE, 2017. 135–145.

[82] Chen L, Fang B, Shang Z, et al. Negative samples reduction in cross-company software defects prediction[J]. Information and Software Technology, 2015, 62:67–77.

[83] Wang C Y, DaghighFarsoodeh A, Pham H V. Selection of Prompt Engineering Techniques for Code Generation through Predicting Code Complexity[J]. arXiv preprint arXiv:2409.16416,

2024..

[84] Li Z, Zhang H, Jing X Y, et al. DSSDPP: Data Selection and Sampling Based Domain Programming Predictor for Cross-Project Defect Prediction[J]. IEEE Transactions on Software Engineer- ing, 2022..

[85] Turhan B, Menzies T, Bener A B, et al. On the relative value of cross-company and within-

company data for defect prediction[J]. Empirical Software Engineering, 2009, 14:540–578.

[86] Ryu D, Jang J I, Baik J. A transfer cost-sensitive boosting approach for cross-project defect

prediction[J]. Software Quality Journal, 2017, 25:235–272.

[87] Ryu D, Choi O, Baik J. Value-cognitive boosting with a support vector machine for cross-project defect prediction[J]. Empirical Software Engineering, 2016, 21:43–71.

[88] Diederik P K. Adam: A method for stochastic optimization[J]. (No Title), 2014..

[89] Scarselli F, Gori M, Tsoi A C, et al. The graph neural network model[J]. IEEE transactions on neural networks, 2008, 20(1):61–80.

[90] Zeng C, Zhou C Y, Lv S K, et al. GCN2defect: Graph Convolutional Networks for

SMOTETomek-based Software Defect Prediction[C]. Proceedings of 2021 IEEE 32nd Inter-

national Symposium on Software Reliability Engineering. IEEE, 2021. 69–79.

[91] Xu J, Wang F, Ai J. Defect prediction with semantics and context features of codes based on graph representation learning[J]. IEEE Transactions on Reliability, 2020, 70(2):613–625.

[92] Liu H, Li Z, Zhang H, et al. CFG2AT: Control Flow Graph and Graph Attention Network-Based Software Defect Prediction[J]. IEEE Transactions on Reliability, 2024..

[93] Brasil-Silva R, Siqueira F L. Metrics to quantify software developer experience: a systematic mapping[C]. Proceedings of Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, 2022. 1562–1569.

[94] Bhattacharya P, Neamtiu I, Faloutsos M. Determining developers’ expertise and role: A graph hierarchy-based approach[C]. Proceedings of 2014 IEEE international conference on software maintenance and evolution. IEEE, 2014. 11–20.

[95] Bergersen G R, Hannay J E, Sjoberg D I, et al. Inferring skill from tests of programming performance: Combining time and quality[C]. Proceedings of 2011 international symposium on empirical software engineering and measurement. IEEE, 2011. 305–314.

[96] Rahman F, Devanbu P. Ownership, experience and defects: a fine-grained study of authorship[C]. Proceedings of Proceedings of the 33rd International Conference on Software Engineering, 2011. 491–500.

[97] Yun S, Jeong M, Kim R, et al. Graph transformer networks[J]. Advances in neural information processing systems, 2019, 32.

[98] Bahaweres R B, Agustian F, Hermadi I, et al. Software defect prediction using neural network based SMOTE[C]. Proceedings of 2020 7th International Conference on Electrical Engineering, Computer Sciences and Informatics (EECSI). IEEE, 2020. 71–76.

[99] Feng S, Keung J, Yu X, et al. Investigation on the stability of SMOTE-based oversampling techniques in software defect prediction[J]. Information and Software Technology, 2021, 139:106662.

[100] Zhao T, Zhang X, Wang S. Graphsmote: Imbalanced node classification on graphs with graph neural networks[C]. Proceedings of Proceedings of the 14th ACM international conference on web search and data mining, 2021. 833–841.

[101] Pearson K. Mathematical contributions to the theory of evolution.—on a form of spurious correlation which may arise when indices are used in the measurement of organs[J]. Proceedings of the royal society of london, 1897, 60(359-367):489–498.