面向物联网软件系统的软件缺陷分配算法研究

Research on Software Defect Allocation Algorithm for IoT Software System

提出了一种基于遗传算法的多样化特征选择和集成极限学习机(ELM)的软件缺陷分配算法,以快速、有效地将缺陷分配给合适的开发人员。首先,从缺陷报告中提取有用的信息。然后对数据进行预处理以建立向量空间模型,并对各种特征选择进行预处理,以选择具有最大统计信息的最小代表性非冗余特征集。最后,设计了集成ELM训练分类器对缺陷报告进行分类。实验结果表明,与现有方法相比,提出的特征选择技术和基于ELM的缺陷分类方法能提高分类的准确性。

This paper proposes a software defect allocation algorithm based on diversified feature selection of genetic algorithm and integrated extreme learning machine(ELM)to quickly and effectively allocate defects to appropriate developers.First,extract useful information from the defect report.Then the data is preprocessed to build a vector space model,and various feature selections are preprocessed to select the smallest representative non-redundant feature set with the largest statistical information.Finally,an integrated ELM training classifier is designed to classify defect reports.The experimental results show that,compared with the existing methods,the feature selection technology and ELM-based defect classification method proposed in this paper can improve the accuracy of classification.