Publication:
Minería de imágenes en repositorios de proyectos de Software soportada por Machine Learning

Thumbnail Image
Files

Files

Primary Rodríguez Torres, Sergio Andrés-2023.pdf (2.39 MB)
 Autorización.pdf (237.58 KB)
View Statistics
Reference Managers
Mendeley
Indexers
Google Scholar CORE
QR Code
QR Code
Authors

Authors

Rodríguez Torres, Sergio Andrés
Abstract (Spanish)
Este documento presenta un análisis enfocado a los diagramas de software en repositorios Git. Para el análisis se construyó un dataset con 5.981 imágenes con las categorías none, diagrama de actividades, diagrama de secuencia, diagrama de clases, diagrama de componentes, diagrama de casos de uso y diagramas cloud. Dicho dataset se usó para el entrenamiento de una red convolucional DenseNet169 pre-entrenada con el dataset ImageNet usando la técnica de transfer learning. La red alcanzó una exactitud en la predicción del 98.6 % y un f1-score de 98.3 %. Luego se usaron técnicas de minería de repositorios para analizar 2’469.206 imágenes equivalentes a 231 GB en datos, obtenidas de 287.201 repositorios. Con el fin de conocer que tan común es cargar imágenes de diagramas, cómo se distribuyen en los repositorios de software y cada cuanto se actualizan.
Abstract (English)
This document presents an analysis focused on software diagrams in Git repositories. For the analysis, a dataset was built with 5,981 images with the categories none, activity diagram, sequence diagram, class diagram, component diagram, use case diagram, and cloud diagrams. The dataset was used to train a DenseNet169 convolutional neural network pre-trained with the ImageNet dataset using the transfer learning technique complemented with fine-tuning. The network achieved a prediction accuracy of 98.6% and an f1-score of 98.3%. Repository mining techniques were then used to analyze 2,469,206 images equivalent to 231 GB of data, obtained from 287,201 repositories. To know how common it is to upload diagram images, their distribution in the repositories, and the update frequency.
Director

Advisors/Directors

Garzón A, Wilmer
Cadavid R, Héctor
Extent
51 páginas
URI: https://repositorio.escuelaing.edu.co/handle/001/2243
Collections

Collections

FD - Trabajos de Grado Maestría en Informática
References

[1] About the Unified Modeling Language Specification Version 2.5 — omg.org. https://www.omg.org/spec/UML/2.5/.

[2] G. Antoniol et al. «Recovering traceability links between code and do- cumentation». En: IEEE Transactions on Software Engineering 28.10 (2002), págs. 970-983. doi: 10.1109/TSE.2002.1041053.

[3] Robert Audi. Cambridge Dictionary of Philosophy. Cambridge Univer- sity Press, 1999.

[4] Thazin Win Win Aung, Huan Huo y Yulei Sui. «A Literature Review of Automatic Traceability Links Recovery for Software Change Impact Analysis». En: Proceedings of the 28th International Conference on Program Comprehension. ICPC ’20. Seoul, Republic of Korea: Associa- tion for Computing Machinery, 2020, págs. 14-24. isbn: 9781450379588. doi: 10.1145/3387904.3389251. url: https://doi.org/10.1145/ 3387904.3389251.

[5] Olusola Tope Babalola. Automatic recognition and interpretation of finite state automata diagrams. Dic. de 2015. url: https://scholar. sun.ac.za/handle/10019.1/97814.

[6] Natalie Best, Jordan Ott y Erik Linstead. «Exploring the efficacy of transfer learning in mining image-based software artifacts». En: Jour- nal Of Big Data 7 (ago. de 2020). doi: 10.1186/s40537-020-00335-4.

[7] Gosala Bethany et al. «Automatic Classification of UML Class Dia- grams Using Deep Learning Technique: Convolutional Neural Network». En: Applied Sciences (mayo de 2021). doi: 10.3390/app11094267.

[8] Francois Chollet. Deep learning with python. en. New York, NY: Man- ning Publications, 2017. isbn: 9781617294433.

[9] Dataset of the Paper .Automatically Classifying UML Class Diagrams from Images using Deep Learning". Zenodo, ene. de 2021. doi: 10. 5281/zenodo.4252890. url: https://doi.org/10.5281/zenodo. 4252890.

[10] Jia Deng et al. «ImageNet: a Large-Scale Hierarchical Image Data- base». En: jun. de 2009, págs. 248-255. doi: 10.1109/CVPR.2009. 5206848.

[11] Arden Dertat. Applied Deep Learning - Part 4: Convolutional neu- ral networks. en. 2017. url: https://towardsdatascience.com/ applied-deep-learning-part-4-convolutional-neural-networks- 584bc134c1e2.

[12] Marta de Esteban Belzuz, Carlos Martin y Francisco Morillo. MPM Construyendo una arquitectura Segura y altamente Disponible en AWS. url: https://aws.amazon.com/es/blogs/aws-spanish/mpm- construyendo-una-arquitectura-segura-y-altamente-disponible- en-aws/.

[13] Georgios Gousios. «The GHTorrent dataset and tool suite». En: Procee- dings of the 10th Working Conference on Mining Software Repositories. MSR ’13. San Francisco, CA, USA: IEEE Press, 2013, págs. 233-236. isbn: 978-1-4673-2936-1. url: http://dl.acm.org/citation.cfm? id=2487085.2487132.

[14] Douglas M. Hawkins. «The problem of overfitting». En: ChemInform 35.19 (2004). doi: 10.1002/chin.200419274.

[15] Regina Hebig et al. «The quest for open source projects that use UML: mining GitHub». En: oct. de 2016, págs. 173-183. doi: 10.1145/ 2976767.2976778.

[16] Gao Huang et al. «Densely Connected Convolutional Networks». En: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Jul. de 2017, págs. 2261-2269. doi: 10.1109/CVPR.2017.243.

[17] B. Karasneh y Michel Chaudron. «Extracting UML models from ima- ges». En: mar. de 2013, págs. 169-178. doi: 10.1109/CSIT.2013. 6588776.

[18] B. Karasneh y Michel Chaudron. «Img2UML: A system for extracting UML models from images». En: sep. de 2013, págs. 134-137. doi: 10. 1109/SEAA.2013.45.

[19] B. Karasneh y Michel Chaudron. «Online Img2UML Repository: An Online Repository for UML Models». En: oct. de 2013.

[20] C. Lanczos. «An iteration method for the solution of the eigenvalue problem of linear differential and integral operators». En: Journal of research of the National Bureau of Standards 45.4 (1950), pág. 255. issn: 0091-0635. doi: 10.6028/jres.045.026. url: http://dx.doi. org/10.6028/jres.045.026.

[21] D. Lu y Q. Weng. «A survey of image classification methods and techniques for improving classification performance». En: Internatio- nal Journal of Remote Sensing 28.5 (2007), págs. 823-870. doi: 10. 1080/01431160600746456. eprint: https://doi.org/10.1080/ 01431160600746456. url: https://doi.org/10.1080/01431160600746456.

[22] Ramírez Luna et al. «Modelo para Almacenar y Recuperar Métri- cas de Software». Español. En: Conciencia Tecnológica (2010). issn: 1405-5597. url: https://www.redalyc.org/articulo.oa?id= 94415753006.

[23] Emmanuel Maggiori et al. «Towards Recovering Architectural Infor- mation from Images of Architectural Diagrams». En: sep. de 2014.

[24] Kaushil Mangaroliya y Het Patel. «Classification of Reverse-Engineered Class Diagram and Forward-Engineered Class Diagram using Machine Learning». En: (nov. de 2020).

[25] Amitha Mathew, Amudha Arul y S. Sivakumari. «Deep Learning Tech- niques: An Overview». En: ene. de 2021, págs. 599-608. isbn: 978-981- 15-3382-2. doi: 10.1007/978-981-15-3383-9_54.

[26] Valentín Moreno et al. «Automatic Classification of Web Images as UML Static Diagrams Using Machine Learning Techniques». En: Ap- plied Sciences 10 (abr. de 2020), pág. 2406. doi: 10.3390/app10072406.

[27] Jordan Ott, Abigail Atchison y Erik Linstead. «Exploring the applica- bility of low-shot learning in mining software repositories». En: Journal Of Big Data 6 (mayo de 2019), pág. 35. doi: 10.1186/s40537-019- 0198-z.

[28] Mr Pankaj S. Parsania y Dr Paresh V. Virparia. «A comparative analy- sis of image interpolation algorithms». En: nternational journal of ad- vanced research in computer and communication engineering 5.1 (2016), págs. 29-34. issn: 2319-5940. doi: 10.17148/ijarcce.2016.5107. url: http://dx.doi.org/10.17148/ijarcce.2016.5107.

[29] Truong Ho-Quang et al. «Automatic Classification of UML Class Dia- grams from Images». En: dic. de 2014. doi: 10.1109/APSEC.2014.65.

[30] George Reese. Cloud application architectures: Building Applications and infrastructure in the cloud. O’Reilly, 2009.

[31] Romain Robbes. «Minería de repositorios de software para ayudar a los desarrolladores». En: Revista Bits de Ciencia 5 (2011), págs. 2-7. issn: 0718-8013. url: https://revistasdex.uchile.cl/index.php/ bits/issue/view/130.

[32] Gregorio Robles et al. «An Extensive Dataset of UML Models in GitHub». En: 2017 IEEE/ACM 14th International Conference on Mining Soft- ware Repositories (MSR). Abr. de 2017, págs. 519-522. doi: 10.1109/ MSR.2017.48.

[33] Sumit Saha. A comprehensive guide to Convolutional Neural Networks - the eli5 way. Nov. de 2022. url: https://towardsdatascience.com/a - comprehensive - guide - to - convolutional - neural - networks - the-eli5-way-3bd2b1164a53.

[34] Selenium: Definition, how it works and Why you need it. Jun. de 2022. url: https://www.browserstack.com/selenium.

[35] Sergei Shcherban et al. «Multiclass Classification of Four Types of UML Diagrams from Images Using Deep Learning». En: mayo de 2021. doi: 10.18293/SEKE2021-185.

[36] Crystal Song. Software architecture diagramming and patterns. Abr. de 2022. url: https://www.educative.io/blog/software-architecture- diagramming-and-patterns.

[37] Farhana Sultana, Abu Sufian y Paramartha Dutta. «Advancements in Image Classification using Convolutional Neural Network». En: 2018 Fourth International Conference on Research in Computational Intelli- gence and Communication Networks (ICRCICN). 2018, págs. 122-129. doi: 10.1109/ICRCICN.2018.8718718.

[38] Warnick Sean West Jeremy Ventura Dan. «A Theoretical Foundation for Inductive Transfer». En: Spring Research (2007).

[39] Rikiya Yamashita et al. «Convolutional neural networks: an overview and application in radiology». En: Insights into Imaging 9 (jun. de 2018). doi: 10.1007/s13244-018-0639-9.