A Study on Improving Adaptive Random Testing

This dissertation, "A Study on Improving Adaptive Random Testing" by Ning, Lareina, Liu, 劉寧, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Abstract of thesis entitled "A Study on Improving Adaptive Random Testing" Submitted by Lareina Ning LIU for the degree of Master of Philosophy at The University of Hong Kong in April 2006 Software testing is an important aspect of software development. Different testing strategies have been proposed in recent years. Random testing is one of thesestrategiesthatselectstestcasesrandomlyfromtheentireinputdomain. It hasbeenfound that byrequiring the testingcases tobe more evenlydistributed, and far separated from each other, the program failure can be more efficiently identifiedforsometypesofdefects. Thisformsthebasisofatestingmethodology called Adaptive Random Testing (referred to as ART). Previous research has classified the patterns of failure-causing inputs into three categories: point, strip and block patterns. Adaptive Random Testing was specially efficient for block failure pattern. To evaluate the failure-finding effi- ciency, researchersproposedanothereffectivenessmetric, theexpectednumberof test cases required to detect the first failure (referred to as F-measure), in addi- tiontothetraditionalP-measureandE-measure. BothDistance-basedAdaptive Random Testing (referred to as D-ART) and Adaptive Random Testing by Re- striction (referred to as R-ART) have been shown to be markedly effective in detecting failure using F-measure as effectiveness metric. Test cases selection in Adaptive Random Testing is an application of Sam- pling, a very important process in statistics which has already generated consid- erable research into sampling efficiency. In this thesis, we review several widely- used sampling methods and use statistic criteria - discrepancy and dispersion tomeasure the distribution of sampling points. We apply such criteria to evaluate ART strategy and compare it with other sampling methods. We propose a different modified Adaptive Random Testing strategy, called Distance Restricted Adaptive Random Testing (DR-ART). It is an extension of Adaptive Random Testing with a combination of both D-ART and R-ART. By adding "farthest away" constraint, it achieves a better control of bad cases, which is the testing process with a F-measure much greater than the average situation. Empirical results showed that DR-ART achieved a better distribution of F-measure than ART. We finally proposed Aging technique and Reset technique to reduce compu- tation overhead in ART. Aging technique is an effective method to lower compu- tation overhead while Reset tries to make use of information of the distribution of F-measure. ----------------- Total words for the abstract: 371 words DOI: 10.5353/th_b3642806 Subjects: Computer software - Testing