An Enhanced ART in High Dimensional Input Domain

Abstract

Random Testing is a commonly used method test data generation. Random testing has the characteristics of high redundancy, low coverage and blindness in generating test cases. Adaptive Random Testing, which puts forward the idea of making the test cases as uniformly distributed as possible, improves the random testing algorithm. Distance-based ART (DART) algorithm is a classical adaptive random testing algorithm, but most of the DART algorithms have boundary effect. With the increase of input domain dimension, the failure detection effectiveness of DART algorithm decreases rapidly. In this paper, we have proposed a novel FSCS algorithm, generating candidate sets through dynamic partitioning in independent dimensions. By constraining the candidate test case generation region, the differences between generated candidate test cases and the generated test cases will be expanded as much as possible, which makes the distribution of test cases more evenly. The simulation results show that the effectiveness of FSCS_AutoP algorithm in high-dimensional input domain space is significantly improved.