A support vector machine based approach for improving accuracy and performance of test oracles / Muhammad Elrashid Yousif Mohamed
One of the key structures in software development is software testing, where there is an increasing need to deal with the issue to provide automated test oracles. Test oracles are simplified and reliable sources that guide testers to undertake a testing process and evaluate faults detected in softwa...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2017
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/14248/ http://studentsrepo.um.edu.my/14248/2/Muhammad_Elrashid.pdf http://studentsrepo.um.edu.my/14248/1/Muhammad_Elrashid.pdf |
| Summary: | One of the key structures in software development is software testing, where there is an increasing need to deal with the issue to provide automated test oracles. Test oracles are simplified and reliable sources that guide testers to undertake a testing process and evaluate faults detected in software. Throughout the years, there has been countless of research conducted on different formats of test oracles, all withholding a similar objective to conclude on whether the test oracle chosen an improvement to software is testing. The objective is to the obtain the challenges currently with existing test oracles and with that identify an approach to address those challenges, using methods such as black-box testing and applying pattern recognizers such just support vector machines as an automated test oracle. In this research, Support Vector Machine, a pattern recognizer based on automated test oracle is introduced to handle mapping and comparison automatically. Previous test oracles, such as Info Fuzzy Networks (IFN) introduced on regression testing contain numerous drawbacks including its limitations to only a single form of testing and its inability to acquire data from other test oracles. Artificial Neural Networks (ANN) based on a single network is a test oracle introduced, in favor of IFN, as it generated accurate data of 91.83% when undertaking testing process. Nevertheless, single network oracle based on ANN has a central drawback and that is its inability to test complex software, making it unreliable. The pattern recognizer SVM is functioned to improve classification performances which will be applied for detecting faults in a testing process. It also contains the capability of solving functional estimated problems and pattern recognition as well as able to undertake complex software. The implication of this research is to provide to software tester different testing methods to apply to projects that may improve accuracy rate as well as reduce cost.
|
|---|