基于Embedded中的Adaboost特征关联数据选择在软件缺陷优化的应用方法研究  

On Ways of Applying Embedded-Based Correlation Data Selection with Adaboost Features in Software Defect Optimization

作　　者：朱小娟[1] Zhu Xiaojuan(Anhui Vocational College of Defense technology,Lu’an 237011,China)

机构地区：[1]安徽国防科技职业学院

出　　处：《黑河学院学报》2019年第4期217-220,共4页Journal of Heihe University

摘　　要：传统Adaboost算法在处理软件缺陷数据时,不能考虑软件缺陷数据为不平衡数据(无缺陷的样本数远远超过有缺陷的样本数)的问题,且通过Adaboost算法选择出来的软件特征之间存在相关性,若不进行及时改进,将影响分类、预测的结果。利用互信息方式,将Adaboost集成算法改进为“MAboost”算法,且改进后算法,具有更强的特征关联数据选择能力。Traditional Adaboost algorithm won’t treat the software defective data as unbalanced data(non-defective samples are much more than defective samples)in dealing with software defect data,thus getting a conclusion that there is correlation between selected software characteristics according to Adaboost algorithm.However,if this practice doesn’t get timely improvement,it will affect the results of the classification and prediction.The Adaboost integration algorithm can changed into the“MAboost”algorithm by means of mutual information,and the improved algorithm has a stronger capability of feature association data selection.

关 键 词：ADABOOST 特征选择 软件缺陷 MAboost 互信息 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]