Stepwise regression clustering method in function points estimation

Abstract

This study proposed a stepwise regression clustering method for software development effort estimation. The proposed algorithm is based on functional points analysis and is used for forming clusters, which contains analogical projects. Furthermore, it is expected that clusters will be shaped well for the regression prediction models. The proposed models are based on Cook distance, which is used for elimination project from clusters. Model performance is proved for selected clusters. Overall model performance influenced by selected clusters, therefore, there is no statistically significant difference between regression models based on clustered and un-clustered datasets.