Fitness Function Design To Improve Evolutionary Structural Testing

Abstract

Evolutionary Structural Testing uses Evolutionary Algorithms (EA) to search for specific test data that provide high structural coverage of the software under test. A necessary characteristic of evolutionary structural testing is that the fitness function is constructed on the basis of the software under test. The fitness function itself is not of interest for the problem; however, a well-constructed fitness function may substantially increase the chance of finding a solution and reaching higher coverage. Better guidance of the search can result in optimizations with less iterations, therefore leading to savings in resource expenditure. This paper presents research results on suggested modifications to the fitness function leading to the improvement of evolutionary testability by achieving higher coverage with less resources. A set of problems and their respective solutions are discussed.