Test Case Selection with Incremental ML

Abstract

Context: Software projects applying continuous integration should run the tests very frequently, but often the number of test is huge and their execution takes a long time. This delays the feedback to the developer. Objective: Study if heuristic and especially incremental machine learning can help in finding an optimal test set that still finds the errors. Method: Several methods for reducing the tests were tested. Each method was applied to the example software its commit history, and the performance of the methods were compared. Results: The test set size can be radically reduced with automatic approaches. Furthermore, it was found that the incremental machine learning based test selection techniques eventually perform equally well or better than the best heuristic.