Construction of an Artificial Intelligence-Based Model for C Language Code Defect Localization

This paper proposes an artificial intelligence-based solution to the problem of C language code defect localization. A systematic analysis of the types and characteristics of C language code defects is conducted. A hybrid model framework integrating machine learning and deep learning is constructed, and various defect localization models are designed and implemented, including random forests, support vector machines, convolutional neural networks, and graph neural networks. Comprehensive experimental evaluations are conducted on common datasets such as PROMISE and CGD, indicating that the proposed models outperform traditional methods in terms of defect localization accuracy, recall, and other metrics, with the highest Fl score reaching 0.91. Additionally, various model fusion and integration strategies are explored to further enhance defect localization performance. The paper also discusses the application scenarios, optimization strategies, and interpretability issues of the models in practical software development, providing new insights for improving software reliability and security.