基于改进GM(1,1)模型预测软件缺陷率  

Model bug rate prediction based on improved GM(1,1)model

作　　者：王曙燕[1] 黄炜青 孙家泽[1] 

机构地区：[1]西安邮电大学计算机学院,陕西西安710121

出　　处：《西安邮电大学学报》2015年第6期69-73,85,共6页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金资助项目(61050003);陕西省教育厅专项科研计划资助项目(11JK1037)

摘　　要：给出一种基于随机抽样一致性算法(RANSAC)的GM(1,1)改进模型。运用RANSAC筛除异常值,选择估计误差最小的点作为定解条件,结合插值思想对软件缺陷率进行预测。分别对有无奇异值的数据预测,结果表明改进后的模型不仅能够改善异常值对预测的影响,而且比GM(1,1)模型取得较高的预测精度。The traditional GM （1,1） model performs poorly in the processing of abnormal data containing singular value. An improved model based on RANdom Sample Consensus （RANSAC） algorithm is therefore proposed. The improved model deselect the singular value via RANSAC algorithm, choose the point which make a minimum estimation error as boundary condition, and combine with interpolation to make the prediction of Model Bug Rate（MBR）. Results show that the improved model can not only improve the effect of the abnormal value but also get higher accuracy than the GM （1,1） model.

关 键 词：灰色模型 随机抽样一致性算法 定解条件 异常值 软件缺陷率 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]