Scholarlink

<p>Predicting software modules prone to faults has become a prominent study focus within software engineering, aiming to spot probable defects early and optimize the allocation of quality assurance efforts. This study proposes a methodology for prediction of fault-prone software modules using a Bayesian Belief Network (BBN). The approach begins by applying information gain-based attribute ranking to a numerical dataset—specifically, the KC1 class-level dataset from the NASA project—categorizing the most effective software metrics. The BBN model is proposed using the top-ranked “Chidamber and Kemerer (CK)” metric suite and a conventional code-size metric. The proposed model is experimented with KC1 data set and validate with previous work. The Comparative evaluation proves that the proposed model achieves better accuracy at 77.93% in fault-prone modules prediction as compared to the previous models that had 67.57% and 75.17%, respectively. The strength of this methodology is based on its systematic integration of information gain attribute ranking, fuzzy reasoning process, and BBN approach, highlighting its effectiveness in advancing fault-prone module 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

Predicting fault-prone software modules using bayesian belief network: an empirical study

  • 期刊论文

  • 发表日期：

    2025年5月7日

查看全文
  • Chandan Kumar,
  • Dilip Kumar Yadav,
  • Mukesh Prasad

摘要

Predicting software modules prone to faults has become a prominent study focus within software engineering, aiming to spot probable defects early and optimize the allocation of quality assurance efforts. This study proposes a methodology for prediction of fault-prone software modules using a Bayesian Belief Network (BBN). The approach begins by applying information gain-based attribute ranking to a numerical dataset—specifically, the KC1 class-level dataset from the NASA project—categorizing the most effective software metrics. The BBN model is proposed using the top-ranked “Chidamber and Kemerer (CK)” metric suite and a conventional code-size metric. The proposed model is experimented with KC1 data set and validate with previous work. The Comparative evaluation proves that the proposed model achieves better accuracy at 77.93% in fault-prone modules prediction as compared to the previous models that had 67.57% and 75.17%, respectively. The strength of this methodology is based on its systematic integration of information gain attribute ranking, fuzzy reasoning process, and BBN approach, highlighting its effectiveness in advancing fault-prone module prediction.