Adaptive Random Test Case Generation Based On Information Entropy

With the development of informatization,computer technology has penetrated into national life and industry.At the same time,software products are widely used in various computing devices,and its scale and complexity are ever-increasing.In this circumstance,software testing has become an important means to ensure quality and reliability.In the field of testing,the selection of test cases has always played a decisive role in guaranteeing the eff-iciency of testing.Therefore,how to pick test cases with better representation has always been an important problem in software testing.Comparing with other methods,adaptive random testing(ART)achieves better effectiveness due to its equally distributed of test cases.Among the existing ART methods,the Fixed-Size-Candidate-Set version of ART(FSCS-ART)exhibits better capability of failure detection,and has been widely used since it was put forward.However,with the increase of input dimensions,the effectiveness of failure detection will deteriorate significantly.What's worse,there are serious computational overhead problems in this algorithm.In view of the above two types of problems,this paper firstly presents the FSCS-Entropy algorithm(FSCS-ART based on Entropy)from the perspective of detection effectiveness,which takes into account the entropy and the shortest distance of candidates to get an equally distribution.Then,from the perspective of improving computational efficiency,a fast algorithm DF-FSCS-Entropy for the low-dimensional case was proposed,as well as two fast algorithms RF-FSCS-Entropy and CR-FSCS-Entropy for high-dimensional case.They reduced the cost of the FSCS-Entropy algorithm by distance-aware forgetting strategy and the basic forgetting strategy respectively.Finally,in order to verify the effectiveness of each algorithm,this paper conducts various experimental tests through simulation research.The experimental results show that,compared with the original FSCS-ART algorithm,the FSCS-Entropy algorithm has a similar low-dimensional detection effect and a stronger failure detection ability in the higher-dimensional input domain.For each algorithm,they greatly improve the efficiency and the detection is slightly less efficient than FSCS-Entropy.Automated test case generation techniques help software developers produce high-quality products in less time,saving significant time and cost through replacing frequently manual tests.In this paper,the strategy for adaptive random testing is improved,and proposed the FSCS-Entropy algorithm and subsequent acceleration algorithm,which has certain significance for the research in academic fields.