Generating Diverse Test Suites for Gson Through Adaptive Fitness Function Selection
Paper i proceeding, 2020

Many fitness functions - such as those targeting test suite diversity—do not yield sufficient feedback to drive test generation. We propose that diversity can instead be improved through adaptive fitness function selection (AFFS), an approach that varies the fitness functions used throughout the generation process in order to strategically increase diversity. We have evaluated our AFFS framework, EvoSuiteFIT, on a set of 18 real faults from Gson, a JSON (de)serialization library. Ultimately, we find that AFFS creates test suites that are more diverse than those created using static fitness functions. We also observe that increased diversity may lead to small improvements in the likelihood of fault detection.

Search-based test generation

fitness function

reinforcement learning


Hussein Almulla

University of South Carolina

Gregory Gay

Göteborgs universitet

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

03029743 (ISSN) 16113349 (eISSN)

Vol. SSBSE 2020 246-252
9783030597610 (ISBN)

Symposium on Search-Based Software Engineering
Bari, Italy,


Annan data- och informationsvetenskap






Mer information

Senast uppdaterat