Adaptive test oracle construction for list processing programs
Date Issued
2010
Date
2010
Author(s)
Chang, Kai-Hsiang
Abstract
Software testing is an important and expensive activity to the software industry, with testing accounting for over 50% of the cost of software. To ease this situation, test automation is very critical to the process of software testing. One important component of the test automation is an executable test oracle. Traditionally an executable test oracle is constructed from the software specifications, software documentation, program assertions or even the program itself, etc.
In this paper, we construct a test oracle from a set of test cases with neural networks. Our approach especially emphasizes on the relation between the input and output lists of the test cases. We also present input/output list relation language (IOLRL) as an auxiliary tool for users to further specify the particular relation between the input and output lists. By mining the knowledge of the input/output relation from the given set of test cases, we construct an executable test oracle with neural networks. This constructed test oracle can be used to automate the software testing. The experiments show our constructed test oracles have good performance.
Subjects
test oracle
neural network
input/output list relation language
test case
black-box testing
software testing
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-99-R97921030-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):de55b5762cd45fe99c1abe688b987341