The effects of duration-based moving windows with estimation by analogy

Abstract

Context: Recent studies have revealed that estimation accuracy can be affected by only using a window of recent projects as training data for building an effort estimation model. The studies also showed that the effect and its extent could be affected by effort estimation methods and windowing policies (fixed size or fixed duration). However, a study of perhaps the most common situation - using Estimation by Analogy (EbA) for effort estimation, and only considering as training data projects completed recently in windows defined by duration - is lacking. Objective: To investigate the effects on estimation accuracy of using the fixed-duration windowing policy, particularly in comparison to fixed-size windows, when using EbA. Method: Using a single-company ISBSG data set studied previously in similar research, we examine the effects of using a fixed-duration windowing policy on the accuracy of estimates using EbA. As a preliminary step, we evaluate the effect of some changes to how we apply EbA itself. Results: Fixed-duration windows can improve the accuracy of estimates with EbA. Some window sizes lead to statistically significant improvements. Reinforcing previous research, the effect is smaller and is seen in a narrower range of window sizes than when fixed-size windows are used. Conclusions: Fixed-duration windows are helpful with this data set when using EbA. Variations in the settings for EbA can change the sizes at which windows are helpful. This suggests the need for reviewing optimal window sizes when adopting a new setting of EbA.