Generating Realistic non-functional Property Attributes for Feature Models

In recent years the analysis of feature models, that contain non-functional properties, via analysis methods is rapidly gaining interest in the field of software product line engineering. Those methods are a useful and reliable tool, when testing is approved under real-world conditions. The majority of methods are, however, still tested with feature models containing random NFP values. Hence, we can not have confidence in the reliability under real-wold circumstances. In this thesis, we present a test suite, which provides the generation of test data, by offering a FM generator, which is capable of creating user adjustable FMs with real-world NFP values. Additionally, we also include feature interactions, which are not considered by other testing suites. Our approach is to use an evolutionary algorithm which optimizes randomly generated NFP values to result in a user selectable real-world NFP distribution. Thereby, testing of analysis methods on FMs with real-world based NFP distributions is made easy and comfortable.