Automated pairwise testing approach based on classification tree modeling and negative selection algorithm
Generating the test cases for analysis is an important activity in software testing to increase the trust level of users. The traditional way to generate test cases is called exhaustive testing. It is infeasible and time consuming because it generates too many numbers of test cases. A combinatorial...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English |
Published: |
2019
|
Subjects: | |
Online Access: | http://eprints.utm.my/id/eprint/98282/1/EasterVivianaSandinMSC2019.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Generating the test cases for analysis is an important activity in software testing to increase the trust level of users. The traditional way to generate test cases is called exhaustive testing. It is infeasible and time consuming because it generates too many numbers of test cases. A combinatorial testing was used to solve the exhaustive testing problem. The popular technique in combinatorial testing is called pairwise testing that involves the interaction of two parameters. Although pairwise testing can cover the exhaustive testing problems, there are several issues that should be considered. First issue is related to modeling of the system under test (SUT) as a preprocess for test case generation as it has yet to be implemented in automated proposed approaches. The second issue is different approaches generate different number of test cases for different covering arrays. These issues showed that there is no one efficient way to find the optimal solution in pairwise testing that would consider the invalid combination or constraint. Therefore, a combination of Classification Tree Method and Negative Selection Algorithm (CTM-NSA) was developed in this research. The CTM approach was revised and enhanced to be used as the automated modeling and NSA approach was developed to optimize the pairwise testing by generate the low number of test cases. The findings showed that the CTM-NSA outperformed the other modeling method in terms of easing the tester and generating a low number of test cases in the small SUT size. Furthermore, it is comparable to the efficient approaches as compared to many of the test case generation approaches in large SUT size as it has good characteristic in detecting the self and non-self-sample. This characteristic occurs during the detection stage of NSA by covering the best combination of values for all parameters and considers the invalid combinations or constraints in order to achieve a hundred percent pairwise testing coverage. In addition, validation of the approach was performed using Statistical Wilcoxon Signed-Rank Test. Based on these findings, CTM-NSA had been shown to be able perform modeling in an automated way and achieve the minimum or a low number of test cases in small SUT size. |
---|