Artificial Neural Network for Automatic Test Oracles Generation

Abstract

Test Oracle is one of the most important problems to be tackled for automatic software testing. Unlike manual test, by reviewing results to determine whether or not result is what it should expected to be, automatic methods only depend on specified decision rules. Artificial Intelligence is a prominent solution for such work. In this research, Artificial Neural Network was proposed to direct oracles automatic generation. The mainly work includes: Heuristic test oracles fit for automatic software testing were introduced firstly. Next, the classifying and predication capabilities of the Artificial Neural Network were analyzed, which is very suitable for some kinds of test oracles generation. Then, a test oracles generator was designed by using ANN model. Lastly, a specified example was given for validation and evaluation. Experiments showed ANN is proper for automatic test oracles generation for some kinds of programs.