The Effect of Moving Windows on Software Effort Estimation: Comparative Study with CART

Abstract

BACKGROUND: Several studies in software effort estimation have found that it can be effective to use a window of recent projects as training data for building an effort estimation model. The previous studies evaluated the use of a window with popular estimation models: linear regression (LR) and estimation by analogy (EbA). Many effort estimation models have been proposed, and the generality of windowing approach still remains uncertain for other effort estimation models, especially for those based on different theory. OBJECTIVE: This study investigates the effect of using a window on estimation accuracy with Classification and regression trees (CART). CART was recently found as a good performance method, and is based on a different theory from LR and EbA. METHOD: We compared the estimation accuracy of a windowing approach and growing approach with the same data set and procedure as the past studies. RESULTS: There is a difference in the estimation accuracy between using a window and not using a window. However, the effctive range of using windows on CART is narrower than that on LR. CONCLUSIONS: Windowing is also effective with CART. However, the range of effectiveness is narrower. The results contribute to the generality of the effectiveness of windowing approach.