A case study of the recursive least squares estimation approach to adaptive testing for software components

Abstract

The strategy used for testing a software system should not be fixed, because as time goes on we may have a better understanding of the software under test. A solution to this problem is to introduce control theory into software testing. We can use adaptive testing where the testing strategy is adjusted on-line by using the data collected during testing. Since the use of software components in software development is increasing, it is now more important than ever to adopt a good strategy for testing software components. In this paper, we use an adaptive testing strategy for testing software components. This strategy (AT/spl I.bar/RLSE/sub c/ with c indicating components) applies a recursive least squares estimation (RLSE) method to estimate parameters such as failure detection rate. It is different from the genetic algorithm-based adaptive testing (AT/spl I.bar/GA) where a genetic algorithm is used for parameter estimation. Experimental data from our case study suggest that the fault detection effectiveness of AT/spl I.bar/RLSE/sub c/ is better than that of AT/spl I.bar/GA and random testing.