Scholarlink

This research introduces a classification framework utilizing deep learning models, including recurrent neural network (RNN), convolutional neural network (CNN), long short-term memory (LSTM), and multilayer perceptron (MLP), to categorize defect reports in software development. The framework employs two word embeddings: word2vec and domain-specific embedding, and evaluates their performance in terms of accuracy, precision, and recall. Evaluation conducted on 7116 defect reports from Redmine and NoSQL databases (Cassandra, HBase, and MongoDB) demonstrates that LSTM outperforms other models, achieving maximum accuracies of 70.2% and 67.3% on the Redmine and NoSQL datasets, respectively.

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 Defect Classification Using Deep Learning Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sushil Kumar,
  • Meera Sharma,
  • S. K. Muttoo,
  • V. B. Singh

摘要

This research introduces a classification framework utilizing deep learning models, including recurrent neural network (RNN), convolutional neural network (CNN), long short-term memory (LSTM), and multilayer perceptron (MLP), to categorize defect reports in software development. The framework employs two word embeddings: word2vec and domain-specific embedding, and evaluates their performance in terms of accuracy, precision, and recall. Evaluation conducted on 7116 defect reports from Redmine and NoSQL databases (Cassandra, HBase, and MongoDB) demonstrates that LSTM outperforms other models, achieving maximum accuracies of 70.2% and 67.3% on the Redmine and NoSQL datasets, respectively.