Improving the Accuracy of Estimating the Probability of Test Case Generation for Simulink Models Using Machine Learning

• Item type: Symposium(1)
• 公開日: 2022-01-28
• タイトル: Improving the Accuracy of Estimating the Probability of Test Case Generation for Simulink Models Using Machine Learning
• タイトル / 言語: en
• タイトル: Improving the Accuracy of Estimating the Probability of Test Case Generation for Simulink Models Using Machine Learning
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: Kyushu University
• 著者所属: Kyushu University
• 著者所属: Kyushu University
• 著者所属: Shibaura Institute of Technology
• 著者所属(英): en
• 著者所属(英): Kyushu University
• 著者所属(英): en
• 著者所属(英): Kyushu University
• 著者所属(英): en
• 著者所属(英): Kyushu University
• 著者所属(英): en
• 著者所属(英): Shibaura Institute of Technology
• 著者名: Tiancheng, Jin; Takuya, Ogata; Yuge, Liu; Kenji, Hisazumi
• 著者名(英): Tiancheng, Jin; Takuya, Ogata; Yuge, Liu; Kenji, Hisazumi
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Model-Based Development (MBD) is gaining popularity in a range of fields. In automatic test case generation, a Simulink model is utilized as an input, and test cases that meet decision criteria such as model coverage are output. As a result, test cases can't be generated for models that don't meet the decision criteria. Furthermore, the test case creation success or failure is frequently unknown until the test case generation is run. Suppose the test case generation success or failure is known before the test case generation. In that case, it is possible to encourage the rewriting of the model differently while keeping the functions being same. Some methods estimate whether a test case can be generated using machine learning, but do not achieve sufficient accuracy. This study proposes a new approach to improve accuracy by adopting a graph neural network. The accuracy for LightGBM and Random Forest is about 75%, and the accuracy of GCN and GAT is 81%. The result shows that there are still possibilities for optimization of the method.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Model-Based Development (MBD) is gaining popularity in a range of fields. In automatic test case generation, a Simulink model is utilized as an input, and test cases that meet decision criteria such as model coverage are output. As a result, test cases can't be generated for models that don't meet the decision criteria. Furthermore, the test case creation success or failure is frequently unknown until the test case generation is run. Suppose the test case generation success or failure is known before the test case generation. In that case, it is possible to encourage the rewriting of the model differently while keeping the functions being same. Some methods estimate whether a test case can be generated using machine learning, but do not achieve sufficient accuracy. This study proposes a new approach to improve accuracy by adopting a graph neural network. The accuracy for LightGBM and Random Forest is about 75%, and the accuracy of GCN and GAT is 81%. The result shows that there are still possibilities for optimization of the method.
• 書誌情報: Proceedings of Asia Pacific Conference on Robot IoT System Development and Platform; 巻 2021, p. 100-101, 発行日 2022-01-28
• 出版者 / 言語: ja
• 出版者: 情報処理学会