Scholarlink

<p>Software bug prediction plays a critical role in ensuring software reliability and reducing maintenance costs. While classical Support Vector Machines (SVMs) are widely used in this domain, their performance can be limited when dealing with complex, high-dimensional, or imbalanced defect datasets. To address these challenges, we propose an Enhanced Quantum Support Vector Classifier (E-QSVC) that integrates entanglement-aware kernel design and adaptive quantum circuit depth to improve predictive performance. Experiments conducted on eight real-world software defect datasets demonstrate that E-QSVC consistently outperforms both classical SVM and standard QSVC models. For instance, in the Jackrabbit and Bitcoin datasets, our model achieves recall improvements exceeding <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(35\%\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(15\%\)</EquationSource> </InlineEquation> over classical SVM, respectively, and up to <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(15\%\)</EquationSource> </InlineEquation> over standard QSVC–a notable advancement in scenarios where undetected defects carry significant cost. These findings underscore the value of quantum entanglement in capturing richer data relationships, establishing E-QSVC as a promising and practical tool for software defect prediction.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Software bug prediction using entanglement-enhanced quantum support vector machines (E-QSVM)

  • 期刊论文

  • 发表日期：

    2025年11月17日

查看全文
  • F. El Ayachi,
  • M. El Baz

摘要

Software bug prediction plays a critical role in ensuring software reliability and reducing maintenance costs. While classical Support Vector Machines (SVMs) are widely used in this domain, their performance can be limited when dealing with complex, high-dimensional, or imbalanced defect datasets. To address these challenges, we propose an Enhanced Quantum Support Vector Classifier (E-QSVC) that integrates entanglement-aware kernel design and adaptive quantum circuit depth to improve predictive performance. Experiments conducted on eight real-world software defect datasets demonstrate that E-QSVC consistently outperforms both classical SVM and standard QSVC models. For instance, in the Jackrabbit and Bitcoin datasets, our model achieves recall improvements exceeding \(35\%\) and \(15\%\) over classical SVM, respectively, and up to \(15\%\) over standard QSVC–a notable advancement in scenarios where undetected defects carry significant cost. These findings underscore the value of quantum entanglement in capturing richer data relationships, establishing E-QSVC as a promising and practical tool for software defect prediction.