EXTRAÇÃO DE CONHECIMENTO A PARTIR DE REGRAS DE ASSOCIAÇÃO ENTRE MÉTRICAS DE CÓDIGO FONTE

• Autor(a) principal: de Almeida, João Luiz Ramalheira
• Data de Publicação: 2017
• Outros Autores: Balancieri, Renato; Universidade Estadual de Maringá, Roecker, Max Naegeler, Leal, Gislaine Camila Lapasini; Universidade Estadual de Maringá, Bermejo, Paulo Henrique; Universidade de Brasília
• Idioma: por
• Título da fonte: Revista de Sistemas e Computação
• Texto Completo: https://revistas.unifacs.br/index.php/rsc/article/view/4717
• topic: Software Engineering; Knowledge engineering in software; Software development process management; Source code metrics

Following and register all the produced artifacts along the software development with the source code metrics and commits messages can be a hard task as the software grows in size and complexity. Data Mining tools, such as the Knowledge Discovery in Database (KDD), can be a helpful resource to detect patterns, characteristics and aspects of the development process and team. This paper presents the use of Association Rules in source code metrics with the goal to extract knowledge of source code repositories to identify important features in software's development. A model based on KDD described and a prototype implementing this model was developed. The prototype is characterized as a primary study relative to the application of the model in an example. This study was conducted aiming to characterize the use of the model in a specific context and serves as proof of concept. Various Apache Foundation’s projects evaluated to extract generalizable patterns of the developers and the impacts in the software product. Based on the outcomes of this tool, project managers can easily identify when the development process is in unwanted way and decide new strategies to put it on the right way. With this, it is concluded that knowledge extraction in source code repositories can be a helpful tool to support the decision-making on the software development.