MULTI-OBJECTIVE COEVOLUTIONARY OPTIMIZATION METHOD FOR MUTATION TESTING
<p align="justify">Mutation Testing is one of the technique to improve error detection capability of test cases. This method suffers a major hindrance that it involves generating a huge number of mutant also all the mutants must be executed to a usually large size of test cases. The...
Saved in:
| Main Author: | |
|---|---|
| Format: | Dissertations |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/29003 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | <p align="justify">Mutation Testing is one of the technique to improve error detection capability of test cases. This method suffers a major hindrance that it involves generating a huge number of mutant also all the mutants must be executed to a usually large size of test cases. The huge size of mutants is actually necessary to ensure it represents a wide scope of error, while a wide range of test case could improve test case coverage, but this would increase processing cost. <br />
<br />
<br />
The implementation of search-based optimization in software engineering optimization problem is recently becoming more popular (search-based software engineering), because it is simple to use, and only need a solution representation and evaluation function definition to implement. <br />
<br />
<br />
Coevolution is one of the search-based approaches that work by divide problem to its parts and optimize it simultaneously. These characteristics seem to be a good match to be used in mutation problem which consists of two related problems: optimization in mutant generation and optimization in test case generation. <br />
<br />
<br />
In this research, we implement coevolution method by integrating mutant generation method (regular-expression-based) with a test case generation method, and genetic algorithm evolution method. Validation is done in two ways, by using the quality approximation value (fitness) and actual solution quality by using a benchmark, the experiment involves six laboratory cases and one real-world case. <br />
<br />
<br />
Validation using fitness value shows that the proposed method is able to improve fitness value of both solution (mutant and test cases) in all cases (laboratory and real cases). Measurement of the actual quality of the solution shows that in term of detected-mutant (DM) and undetected-mutant (UM) level, the proposed method is able to reach 95% for DM and 96% for UM compared to the benchmark, and this with a 88% smaller test-case size. In four cases are able to reach the benchmark quality (100% DM and 100% UM), one case exceeds benchmark quality (105% UM), and real case reaches 96% DM.<p align="justify"> <br />
|
|---|