Avaliação Experimental da Detecção Interativa de Anomalias de Código

• Ano de defesa: 2023
• Autor(a) principal: Albuquerque, Danyllo Wagner
• Orientador(a): Almeida, Hyggo Oliveria
• Banca de defesa: Santos, Danilo Freire de Souza , Costa, Evandro de Barros , Silva, Lenardo Chaves , De Oliveira, Roberto Felicio
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• dARK ID: ark:/80033/0013000001pwz
• Idioma: por
• Instituição de defesa: Universidade Federal de Campina Grande
• Programa de Pós-Graduação: Programa de pós-graduação em computação
• Departamento: Universidade Federal de Campina Grande
• País: Brasil
• Palavras-chave em Português: Engenharia de Software; Anomalias de Código; Detecção Interativa; Avaliação Empírica.; Qualidade de Software
• Área do conhecimento CNPq: Sistemas de computação
• Link de acesso: https://deposita.ibict.br/handle/deposita/410

Code smells generally indicate the presence of deeper problems in the software, making its maintenance and evolution difficult. Several smells are cataloged in the literature, and their detection is traditionally supported by Non-Interactive Detection (NID) approaches. These approaches do not support the progressive interaction of developers with the affected code, revealing occurrences of more global smells only at the developer’s request, recurrently implying a late identification of these smells. With the emergence of Interactive Detection (ID), the aim is to deal with the limitations of traditional approaches, allowing the revelation of instances of code smells without an explicit request from the developer, encouraging early detection as a recommended practice. Although developers consider using ID approaches, guidelines were not found in the literature on when and how such approaches should be used in the context of software development activities. In this work, the objective is the experimental evaluation of the interactive detection of code smells in the context of the activities of the software development process. For doing so, studies were conducted to identify the need to use such an approach, evaluate adherent detection methods, develop automated support with DI characteristics, present empirical evidence on its effectiveness in detecting anomalies, as well as describe a way of integration to the development process according to the Scrum framework. The experimental tasks revealed that using the ID technique led to an increase of up to 40% in the recall and up to 25% in the precision in the detection of code smells during inspection activities and code development. Based on the results of the studies, it is concluded that factors associated with ID contributed to the early detection of a greater number of occurrences of code smells compared to traditional approaches. Consequently, using the ID approach in a disciplined way in a development process can promote continuous evaluation and improve the quality of the software.