Analysis of adaptive random testing efficiency based on the experiment

Adaptive random testing (ART) had been proposed as an enhancement of random testing with some algorithms, thus the ability of failure detection was improved. However, previous studies emphasised on the advantages of F-measure a lot but rarely considered the real testing environment. Three typical adaptive random testing algorithms and random testing algorithm were selected, and the impact of the compactness of failure regions, dimensions on the efficiency of ART was analyzed. Then fixed size candidate set (FSCS), which was the best ART algorithm, was compared with random testing in the non-numeric conditions. The results show that the range of application of ART is limited by many factors of input domain. And in real test, the enhancement of test efficiency is not obvious.