Author:
Martins Mariana Crisostomo,Kudo Taciana Novo,Bulcão-Neto Renato F.
Abstract
In recent years, there has been a focus shift from software development in general to the construction and training of machine learning (ML) models integrated into a software product. This movement has raised challenges in ML systems’ requirements engineering (RE) theory and practice. This paper investigates RE practices in ML systems research, development, and innovation projects carried out by an Artificial Intelligence (AI) Unit of the Brazilian Industrial Research and Innovation Company. Our methodology includes semi-structured interviews with leaders of 21 projects and data analysis through the grounded theory method. We identified the predominance of RE methods, techniques, and tools applied ad hoc and uncoordinatedly. This result corroborates the literature reports on RE for ML systems, especially those involving innovation projects.
Publisher
Sociedade Brasileira de Computação
Reference14 articles.
1. Alves, A. P. S. et al. (2024). Status quo and problems of requirements engineering for machine learning: Results from an international survey. In Kadgien, R., Jedlitschka, A., Janes, A., Lenarduzzi, V., and Li, X., editors, Product-Focused Software Process Improvement, pages 159–174, Cham. Springer Nature Switzerland.
2. Chemuturi, M. (2013). Requirements Engineering and Management for Software Development Projects. Springer New York, 1st edition.
3. Corbin, J. and Strauss, A. (2014). Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory. SAGE Publications, 4th edition.
4. Dunbar, B. (2017). National aeronautics and space administration (nasa). technology readiness level. [link].
5. Glaser, B. G. (1992). Basics of grounded theory analysis: Emergence vs forcing. Sociology Pr.