Centroid Based Adaptive Random Testing for object oriented program

A variant of Adaptive Random Testing (ART) called Centroid Based Adaptive Random Testing (CB-ART) has been developed. It aims to improve efficiency of ART which was claimed to have better performance than Random Testing (RT) in finding first error in a program. CB-ART is applying the concept of centroid for test case selection process. The technique for generating and computing centroid from a pool of input objects is the novelty of CB-ART. The experiment results show that our approach can generate 792.38% (near 8 times) more test cases than ARTGen and 108% more test cases than RANDOOP. Although capable in producing more test cases, the effectiveness of CB-ART is equal or higher than ARTGen and RANDOOP in some cases. We claim that our technique is able to generate high quality test case with shorter generation time.