Arselda: An Improvement on Adaptive Random Testing by Adaptive Region Selection

Abstract

Distance-aware Forgetting Fixed Size Candidate Set (DF-FSCS) is an Adaptive Random Testing (ART) technique, which lowers the computational overhead of Fixed Size Candidate Set ART (FSCS-ART), using a forgetting strategy. DF-FSCS partitions the input domain into regions, and while computing the distance of a candidate test case from executed test cases, as a vector in the input domain, it only considers test cases that are in the same region as the candidate. Although being a lightweight technique, there are two issues with DF-FSCS. First, it does not attempt to generate test cases in low-density regions, which if done, could result in a more even spread of test cases. Second, the regions it defines are smaller at the lower or upper boundaries of input domains, which declines the quality of test cases produced in these regions. We propose Arselda, an APR technique that improves DF-FSCS. By generating test cases in low-density regions that have a fewer number of test cases and enlarging regions at lower or upper boundaries of input domains, Arselda addresses the two issues mentioned above. Considering DF-FSCS as the baseline, a simulation analysis has been performed to evaluate the effectiveness of Arselda. According to the experiment results, Arselda has better failure detection effectiveness compared with the baseline for the block failure pattern. Also, Arselda has lower computational overhead than the baseline.