Machine Intelligence in Automated Performance Test Analysis

Abstract

Software testing is a large part of development and especially important for projects that practice Continuous Integration. In order to reduce the burden of testing and make the process more efficient, as much as possible is automated. In this thesis, a design science approach is used to investigate how machine intelligence can be used to improve the automation of the analysis of non-functional testing. A prototype is created in order to demonstrate the ability of machine intelligence methods to provide useful information about the relationships between different test cases and their histories. This prototype was found to be fairly accurate in its predictions of test results, could identify most related degradations across test cases, and was positively received by the testers. Based on the results of this thesis, machine intelligence was found to have great potential in dealing with the large amount of data created during testing.