Application And Improvement Of Capture-Recapture Method In Software Inspection

As an important part of software quality assurance, software inspection has been paid more and more attention in recent years. Software review can find defects in earlier software project and solve problems with less cost. As one of the assistant techniques in software inspection, defect prediction can effectively help determine the status of defects and analyze the necessity of further inspections. Capture-recapture is an estimation method derived from Biodemography, estimating the polulation based on distribution rather than historical data. We can estimate total defects number through calculating the relation matrix of inspectors and defects. Variety of models and estimators has been studied and proved to estimating total defects accurately.Our study evaluated three accurate estimator, Mt-Ch, Mh-Ch and Mh-JK. We used the mean value of them for practical estimation. We also have used co-efficient of covariation (CCV), and improve it to quasi co-efficient of covariation (Q-CCV) as an actual available method to wipe out the dependence of inspectors, for lower relative error value and better estimation performance. These improvements above has been realized in Review Online System, as a completion of capture-recapture method applied in software inspection, supporting companies to use capture-recapture method for defect estimation during software inspection in software development, and building a better software project quality assurance.