A genetic algorithm test generator
Use of a genetic algorithm and formal concept analysis to generate test data for branch coverage is explored in a prototype automatic test generator (ATG) called genet . genet is unique in the sense that it requires minimal source code instrumentation and analysis, and is programming language indepe...
| Main Author: | |
|---|---|
| Format: | Others |
| Published: |
2004
|
| Online Access: | http://spectrum.library.concordia.ca/8112/1/MQ94745.pdf Khor, Susan Lay Choo <http://spectrum.library.concordia.ca/view/creators/Khor=3ASusan_Lay_Choo=3A=3A.html> (2004) A genetic algorithm test generator. Masters thesis, Concordia University. |
| Summary: | Use of a genetic algorithm and formal concept analysis to generate test data for branch coverage is explored in a prototype automatic test generator (ATG) called genet . genet is unique in the sense that it requires minimal source code instrumentation and analysis, and is programming language independent. Besides the novelty of using formal concept analysis within a genetic algorithm, genet extends the opportunism of another evolutionary ATG. Experiments were designed to evaluate the effectiveness of genet and the importance of selection in the evolution of test data. The results of the experiments indicate genet is most effective when selection plays a significant role. This is the case when test solutions for a program are necessarily organized. When it is not necessary for test solutions to resemble each other, adaptation appears to be the more dominant factor and the identification of suitable genetic operators becomes more important. Nevertheless, even in the latter situation, the presence of genet accelerated the evolutionary process for our test programs. Notwithstanding equal adaptation instructions, genetics mattered. |
|---|