| One of the important issues in software testing is to provide an automated test oracle. Test oracles are reliable sources of how the software under test must operate. In the generation of an automated oracle, three challenges were identified i.e. output domain generation, input domain to output domain mapping, and a comparator to decide on the accuracy of the actual outputs. The aim of this research is to propose an automated test oracle framework that addresses these challenges. In particular, I/O Relationship Analysis is proposed to generate the output domain automatically, and Multi-Networks Oracles based on Artificial Neural Networks are introduced to handle the mapping challenge. The last challenge is addressed using an automated comparator that adjusts the oracle precision by defining the comparison tolerance. The proposed framework was evaluated using two case studies. The quality of the proposed oracle was measured by assessing its accuracy, precision, misclassification error, practicality and usability using mutation testing. In addition, Single-Network Oracles were also provided in order to highlight the superiority of the proposed Multi-Networks Oracle. Similarly, a fully automated test driver is provided to execute and evaluate the test cases using the proposed oracle model. Finally, a comparative study between the prominent oracles and the proposed one is provided based on how they solve the oracle challenges and the degree of automation they provide. The results of the study show the proposed frameworks may automate the oracle generation process up to 97.5% with accuracy up to 98.93%. Moreover, the quality of the Multi-Networks Oracle was higher than the Single-Network one in all of the conducted experiments. |
|
Date: Wednesday, November 22, 2017 Language: English Downloded 19 times. |