Replicated analyses of windowing approach with single company datasets
Sousuke Amasaki · 2011
In effort estimation model construction, it seems effective to window training project data so that recently finished projects are only used. The past study examined this windowing approach with ISBSG R10 Data. However, this approach has not been validated with single company dataset. Aim: To investigate effects of windowing approach at a company and generality of observations in the past study. Method: We replicated the past study with two other datasets: CSC and Maxwell datasets. Results: Windowing approach improved predictive performance. However, it is practically useful only for Maxwell dataset. Conclusions: This result contributes to understand the effects of windowing approach under practical situation.