A Discrete Dynamic Artificial Bee Colony with Hyper-Scout for RESTful web service API test suite generation


ŞAHİN Ö., AKAY B.

Applied Soft Computing, vol.104, 2021 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 104
  • Publication Date: 2021
  • Doi Number: 10.1016/j.asoc.2021.107246
  • Journal Name: Applied Soft Computing
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Applied Science & Technology Source, Compendex, Computer & Applied Sciences, INSPEC
  • Keywords: Test suite generation, RESTful, Artificial bee colony algorithm, OPTIMIZATION
  • Erciyes University Affiliated: Yes

Abstract

© 2021 Elsevier B.V.Microservices have become quite important because they decompose an application into a collection of highly maintainable services and reduce the complexity of monolithic architecture. Although the most common method of microservice development is the REST architecture, the number of studies on the automatic testing of RESTful APIs are very limited. For automatic testing of RESTful APIs, different algorithms have been proposed. However, their performance depends on the type of problem being addressed. This study implements the basic Artificial Bee Colony (ABC) algorithm for automatic testing of RESTful web services, and addresses some shortcomings for the RESTful test generation. To overcome these shortcomings, we propose a Discrete Dynamic Artificial Bee Colony with Hyper-Scout (DABC-HS) algorithm for the service-oriented problem. We enhanced the exploration ability of the algorithm by integrating a hyper-scout unit. The presence of solutions that correspond to covered targets but are not related to the remaining targets causes overburdening of the search. In the proposed approach, the onlooker bee phase generates a pool of all solutions and their mutants and applies a dominance-based selection operator to select high-quality solutions related to uncovered targets. The experimental results demonstrate that the DABC-HS recorded the highest performance in four of the seven problems while demonstrating similarly good performance in other problems. MOSA, which is the powerful algorithm in the literature, was the best in two of seven problems while ABC algorithm achieved the best performance in one problem. Although the method could not achieve the best result in all problems, it is more successful than the other published methods in the literature.