Dealing with missing software project data

摘要: Whilst there is a general consensus that quantitative approaches are an important part of successful software project management, has been relatively little research into many the obstacles to data collection and analysis in real world. One feature characterises sets we deal with missing or highly questionable values. Naturally this problem not unique engineering, so explore application two existing imputation techniques have used good effect elsewhere. In order assess potential value use industrial sets. Both quite problematic from effort modelling perspective because they contain few cases, significant number values projects heterogeneous. We examine quality fit models derived by stepwise regression on raw imputed various compared. both find k-nearest neighbour (k-NN) sample mean (SMI) significantly improve model fit, k-NN giving best results. These results consistent other recently published results, consequently conclude can assist empirical engineering.