Scholarlink

<p>In mobile app development, quickly identifying and fixing bugs is crucial for maintaining a seamless user experience and supporting an expanding user base. The significance of this has led to the prominence of Just-in-Time Defect Prediction (JITDP) as a proactive strategy. Existing JITDP approaches in mobile applications focus on improving prediction accuracy at the commit level, hence the need to enhance defect prediction at a fine-grained line-of-code (LOC) level persists. This paper uniquely concentrates on constructing a JITDP model for mobile applications, utilizing mobile domain-relevant change metrics to predict defects at the line level within commits. Our proposed approach, FLAME-JIT (Fine-grained LocalizAtion of Application Mobile dEfects at Just-in-time), leverages Random Forest, Logistic Regression, and Naive Bayes classifiers for buggy commit identification and integrates the N-gram language model for localization of defects at the LOC level. We evaluated FLAME-JIT on 13 open-source mobile applications and achieved 80% accuracy in detecting defective commits and a substantial 64% Top10 accuracy in locating defective lines. This provides a foundation for further refinement of fine-grained JITDP models in mobile app development and enables practitioners to conduct more effective inspections.</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

FLAME-JIT: Fine-grained localization of application mobile defects at just-in-time

  • 期刊论文

  • 发表日期：

    2025年12月23日

查看全文
  • Itrooba Marrium,
  • Javaria Imtiaz,
  • Nigar Azhar Butt

摘要

In mobile app development, quickly identifying and fixing bugs is crucial for maintaining a seamless user experience and supporting an expanding user base. The significance of this has led to the prominence of Just-in-Time Defect Prediction (JITDP) as a proactive strategy. Existing JITDP approaches in mobile applications focus on improving prediction accuracy at the commit level, hence the need to enhance defect prediction at a fine-grained line-of-code (LOC) level persists. This paper uniquely concentrates on constructing a JITDP model for mobile applications, utilizing mobile domain-relevant change metrics to predict defects at the line level within commits. Our proposed approach, FLAME-JIT (Fine-grained LocalizAtion of Application Mobile dEfects at Just-in-time), leverages Random Forest, Logistic Regression, and Naive Bayes classifiers for buggy commit identification and integrates the N-gram language model for localization of defects at the LOC level. We evaluated FLAME-JIT on 13 open-source mobile applications and achieved 80% accuracy in detecting defective commits and a substantial 64% Top10 accuracy in locating defective lines. This provides a foundation for further refinement of fine-grained JITDP models in mobile app development and enables practitioners to conduct more effective inspections.