ReqCluster4IoT: um método de agrupamento de requisitos para aplicações de IoT

• Ano de defesa: 2025
• Autor(a) principal: SILVA, Bruno Carvalho da
• Orientador(a): SANTOS, Davi Viana dos
• Banca de defesa: SANTOS, Davi Viana dos , SANTOS, Rodrigo Pereira dos, BRAZ JUNIOR, Geraldo, MALCHER, Paulo Robson Campelo, CONTE, Tayana Uchôa
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Maranhão
• Programa de Pós-Graduação: PROGRAMA DE PÓS-GRADUAÇÃO EM CIÊNCIA DA COMPUTAÇÃO/CCET
• Departamento: DEPARTAMENTO DE INFORMÁTICA/CCET
• País: Brasil
• Palavras-chave em Português: Engenharia de Requisitos; Internet das Coisas; Processamento de Linguagem Natural; Bases de Requisitos; Agrupamento de Requisitos.
• Palavras-chave em Inglês: Requirements Engineering; Internet of Things; Natural Language Processing; Requirements Bases; Requirements Clustering.
• Área do conhecimento CNPq: Ciência da Computação
• Link de acesso: https://tedebc.ufma.br/jspui/handle/tede/6220

Requirements Engineering is a critical phase in the software development lifecycle, ensuring the nal product meets stakeholder needs. In software projects, teams often write requirements in unstructured natural language and organize them into requirement specication documents. While this approach simplies document creation, it makes manual analysis dicult and time-consuming, especially in projects with numerous requirements. Modern software contexts, such as the Internet of Things (IoT), demand continuous reevaluation of software development processes. Requirement analysis in IoT poses challenges due to the high diversity of elements in this complex domain. In such scenarios, clustering algorithms can support requirement analysis by grouping them into similar contexts, improving system understanding. To address this, we propose ReqCluster4IoT, a functional requirement clustering method for IoT applications. The approach builds on an exploratory literature analysis to identify IoT characteristics, along with experiments on requirement classication and text semantic similarity. To enhance classication and semantic similarity computation, we developed two datasets: Promise+ and ReqFuncimDataset. Experiments showed that Transformer-based models outperform traditional approaches. We evaluated the method’s acceptance and usefulness through a survey and a focus group. The survey gathered feedback from 10 IoT software developers, while the focus group provided deeper insights into the method’s results. Findings conrmed that ReqCluster4IoT eectively captures relationships between requirements for accurate grouping. Additionally, the method helps professionals cluster IoT software requirements and identify IoT-specic features within groups, improving requirement analysis. The method achieved high acceptance rates in terms of usefulness, eort reduction potential, and intention to use.