Normalized Adaptive Random Test for Integration Tests

• Published in: 2010 IEEE 34th Annual Computer Software and Applications Conference Workshops pp. 335 - 340
• Main Authors: Shin, Seung-Hun, Park, Seung-Kyu, Choi, Kyung-Hee, Jung, Ki-Hyun
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2010
• Subjects: Adaptive Random Testing; Closed box; Distortion; Integration Testing; Partitioning algorithms; Random Test; Software; Software algorithms; Software development management; Software measurement; Strips; Subspace constraints; Test case selection; Testing
• ISBN: 1424480892; 9781424480890
• DOI: 10.1109/COMPSACW.2010.65

The Adaptive Random Testing (ART) was devised to improve the performance of pure random tests, which is one of black-box testing strategies. The ART-based algorithms were developed mainly for unit or single module tests. When a given unit-under-test (UUT) is integrated with an already proven front-end software module which takes inputs and supplies the outputs to the UUT, the performance of ART-based algorithm applied to the integrated software is severely degraded depending on the behavior of front-end software. In this paper, a normalized ART-based algorithm is proposed for the integration and regression tests where an UUT is integrated with a front-end software module. The front-end software with three different functions, Log, Exponential, and Normal function, is experimented by the simulation to show the performance of the proposed method. Depending on the skewness driven by the function of front-end, the experimental results show that the proposed method outperforms significantly the ART without normalization in terms of F-measure.