Scholarlink

<p>The IVY algorithm is a novel heuristic optimization algorithm. In the later iteration of the IVY algorithm, it is easy to fall into the local optimal solution. Therefore, we proposed the Bagging Gradient Nearest Neighbor Ivy algorithm (BGNIVY). This algorithm enhances the global exploration and local development capabilities by combining the Bagging lens imaging inverse learning and the gradient nearest neighbor search, breaking through the limitations of the traditional Ivy algorithm in the optimization process. In addition, in the problem of cross-project software defect prediction, traditional support vector machines (SVM) face the problems of difficult parameter tuning and low prediction accuracy. We use the <b>B</b>GN<b>I</b>VY algorithm to optimize <b>S</b>VM and use the optimized SVM as the base classifier to construct an <b>E</b>nsemble <b>L</b>earning for heterogeneous software defect prediction (BISEL). Comparative experiments on eight benchmark test functions demonstrate that BGNIVY achieves superior optimization accuracy, with Wilcoxon test results confirming statistically significant performance improvements. Subsequently, BGNIVY is employed to optimize the key parameters of SVM, which is then combined with Bagging ensemble learning to construct the BISEL algorithm for heterogeneous defect prediction. Evaluating on 34 software defect datasets from five distinct project categories reveals that BISEL significantly outperforms six other comparative algorithms across four evaluation metrics, indicating higher prediction accuracy and stronger generalization capabilities. Furthermore, statistical analysis based on Friedman ranking and Nemenyi post-hoc tests shows notable variations between BISEL and the other algorithms.</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

Bagging gradient nearest neighbor-based Ivy algorithm optimized SVM ensemble learning for heterogeneous software defect prediction

  • 期刊论文

  • 发表日期：

    2025年12月29日

查看全文
  • Li-fang Chen,
  • Ke-xin Cao,
  • Si-peng Zhang,
  • Qi Dai

摘要

The IVY algorithm is a novel heuristic optimization algorithm. In the later iteration of the IVY algorithm, it is easy to fall into the local optimal solution. Therefore, we proposed the Bagging Gradient Nearest Neighbor Ivy algorithm (BGNIVY). This algorithm enhances the global exploration and local development capabilities by combining the Bagging lens imaging inverse learning and the gradient nearest neighbor search, breaking through the limitations of the traditional Ivy algorithm in the optimization process. In addition, in the problem of cross-project software defect prediction, traditional support vector machines (SVM) face the problems of difficult parameter tuning and low prediction accuracy. We use the BGNIVY algorithm to optimize SVM and use the optimized SVM as the base classifier to construct an Ensemble Learning for heterogeneous software defect prediction (BISEL). Comparative experiments on eight benchmark test functions demonstrate that BGNIVY achieves superior optimization accuracy, with Wilcoxon test results confirming statistically significant performance improvements. Subsequently, BGNIVY is employed to optimize the key parameters of SVM, which is then combined with Bagging ensemble learning to construct the BISEL algorithm for heterogeneous defect prediction. Evaluating on 34 software defect datasets from five distinct project categories reveals that BISEL significantly outperforms six other comparative algorithms across four evaluation metrics, indicating higher prediction accuracy and stronger generalization capabilities. Furthermore, statistical analysis based on Friedman ranking and Nemenyi post-hoc tests shows notable variations between BISEL and the other algorithms.