Abstract: Because capture-recapture’s estimators have a general trend of underestimation, this paper puts forward a new experience-based algorithm. Based on some data of two literatures, virtual software inspections are carried out, and the performance of both original and improved estimators are compared and analyzed. Result verifies that the estimation accuracy of new estimators is better than that of the olds.

Key words: software inspection, defect content estimation, capture-recapture

摘要： 针对捕获-再捕获模型的估计器普遍存在低估的现象，提出一种基于历史数据的缺陷估计改进方法，在基于已有研究的部分数据进行虚拟评审后，对比了改进前后估计器的缺陷估计效果，结果表明，改进后的估计器普遍较改进前的估计器的估计精度要高。

关键词: 软件评审, 缺陷估计, 捕获-再捕获

CLC Number: 

• TP393