Development of artificial intelligence based software testing platform

Due to the fast-paced development of technology, the software has become a crucial aspect of modern life, faciliating the operation and management of hardware devices. Faulty and low-quality software can pose severe risks to individuals and endanger their lives. This highlights the necessity of developing error-free and high-quality software. Verification and validation processes are integral to achieving error-free and high-quality software, with software testing being a fundamental part of this process. While code coverage is a widely used software testing technique to assess the effectiveness of test suites, it has certain limitations. The mutation testing approach has been proposed in industry and academic research to address these limitations. Mutation testing is considered a method for generating test scenarios and evaluating the effectiveness of test suites. This thesis aims to automatically learn mutations from real-world bugs in software applications and present them on a platform. The proposed method involves initially extracting bug fixes at the method-level, categorizing them according to mutation types, and abstracting the source code. Subsequently, a neural machine translation-based transformer model has been developed to generate mutant code from the abstracted method-level source code based on the mutation types. Finally, a platform has been created for conducting software testing, including the proposed neural machine translation-based mutant generation model. The experimental results demonstrate that the proposed method successfully predicts mutation codes similar to real-world mutations, ranging from 6% to 35%. Additionally, the mutant generator model can successfully produce mutant codes different from the original buggy source code with a success rate of 94%. The generated mutant source codes can be obtained with 96% accuracy regarding lexical and syntactic correctness. These results indicate that source codes can be artificially generated for software testing based on artificial intelligence.