Aplicação de Data Analytics em dados retrospectivos de coberturas de código em testes automatizados

Abstract

Automated software testing can prevent numerous software defects in the early stages of development. Among the tools that do automated testing there are code coverage libraries, which instrumenting the source code, allowing the developer to know exactly the points that have not yet been executed and highlighting those that need further verification as well. These tools produce a series of data that can be analyzed, creating the possibility to store and analyze the data. So using a non-parametric statistical approach, the study verifies developer profile data and code coverage data produced by libraries, to find the features that best appoint developers that need training in automated software testing. The results show characteristics from low testing code coverage profiles. It’s point to possible groups that need more focus on internal trainings.