Estimation of Software Integration Test Duration via UML Statecharts

In integration testing it can be very useful to predict how much time it will take to complete a set of integration test runs, especially during the version changes of software projects. Past studies usually focus on estimating the whole test suite effort rather than the test delta required to be applied during version changes which can be expected to increase the margin of error during the estimation process. In this study we apply well-knwon prediction techniques over software projects that make use of Model Driven Engineering (MDE) and UML-statecharts per se, in order to estimate the duration of a single run of integration tests during version changes. In particular, we apply a linear regression model to estimate the exact duration and Support Vector Machine (SVM) classifiers to predict the duration by using three categories: 'short', 'medium' and 'long'. Our results give important hints about how to perform such analysis in the chosen domain.