• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International Conference on Cyber and IT Service Management >Combining integreted sampling technique with feature selection for software defect prediction
【24h】

Combining integreted sampling technique with feature selection for software defect prediction

机译：结合集成采样技术和特征选择进行软件缺陷预测

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Good quality software is a supporting factor that is important in any line of work in of society. But the software component defective or damaged resulting in reduced performance of the work, and can increase the cost of development and maintenance. An accurate prediction on software module prone defects as part of efforts to reduce the increasing cost of development and maintenance of software. An accurate prediction on software module prone defects as part of efforts to reduce the increasing cost of development and maintenance of software. From the results of these studies are known, there are two problems that can decrease performance prediction of classifiers such imbalances in the distribution of the class and irrelevant of the attributes that exist in the dataset. So as to handle both of these issues, we conducted this research using integrated a sample technique with feature selection method. Based on research done previously, there are two methods of samples including random under sampling and SMOTE for random over sampling. While on feature selection method such as chi square, information gain and relief methods. After doing the research process, integration SMOTE technique with relief method used on Naïve Bayes classifiers, the result of the predicted value better than any other method that is 82%.
机译：高质量的软件是对社会上任何工作都至关重要的支持因素。但是软件组件有缺陷或损坏会导致工作性能降低，并可能增加开发和维护成本。对软件模块容易产生缺陷的准确预测是减少软件开发和维护成本增加的努力的一部分。对软件模块容易产生缺陷的准确预测是减少软件开发和维护成本增加的努力的一部分。从这些研究的结果知道，存在两个问题可能会降低分类器的性能预测，例如类的分布不平衡和数据集中存在的属性的不相关。为了解决这两个问题，我们将样本技术与特征选择方法集成在一起进行了这项研究。根据以前的研究，有两种采样方法，包括随机欠采样和随机过采样的SMOTE。在使用特征选择方法（例如卡方），信息获取和缓解方法时。在完成研究过程后，将朴素贝叶斯分类器上使用的SMOTE技术与缓解方法相集成，预测值的结果要好于其他任何方法（82％）。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号