Identifying inconsistent software defect predictions with symmetry metamorphic relation pattern.
Abstract
syntax trees of source code and features provided by the PROMISE repository to predict defects (Wang et al., 2021), unsupervised software defect prediction approach used threshold derivation for clustering to label the data points for defect prediction (Kumar et al., 2022). Ensemble different machine learning techniques and attention mechanisms to predict software defects can also provide more detailed information on the location of defects in code (Zhang et al., 2020). Cross-project defect prediction method (CPDP) is another software defect prediction approach that uses data from other 2.3. Explainable artificial intelligence (XAI) Although MT is a valuable tool in validating machine learning models, understanding the model's decision-making process is essential (Molnar, 2020). Despite applications of machine learning algorithms that have been implemented in various areas, there is still significant potential for improvement, and understanding the basis for the final decisions made by these models remains challenging (Lin et al., 2024; Zheng et al., 2022). This lack of transparency complicates the estab lishment of trust between humans and machines (Ajorloo et al., 2024; Lin et al., 2024; Zheng et al., 2022). In addition, the 2022 program so licitation "NSF22-502: National Artificial Intelligence (AI) Research Institutes: Accelerating Research, Transforming Society, and Growing the American Workforce" of U.S. National Science Foundation (NSF) has stated that AI system is trustworthy as "beyond reliability and 2