Assessing impacts of economic and municipality factors on building construction duration using ensemble categorical boosting

Peiman F.; Khalilzadeh M.; Shahsavari-Pour N.; Ravanshadnia M. · Discover Applied Sciences (2025) · DOI: 10.1007/s42452-025-07078-y

Resumen

In addition to the need to increase the accuracy of models, there is a lack of consensus among researchers about considering factors of external economic in duration prediction models. First, the categorical boosting (CatBoost-2018) model in the first-level was created aiming to find the optimal combination of inputs with the best results for the duration. In continuation, only the second-level model i.e. Stacking-CatBoost2 model was created aiming to noticeably increase the estimation accuracy of two models of the first-level including the artificial neural network (ANN) single model and CatBoost1. The final stacking results also have shown the superiority of that model over three other study models according to the K-fold method (used for validation and verification). Then, the results of sensitivity analysis (SA) using the error deterioration measurement method have shown the negative impact of an economic input and the significant positive impact of a unique input to the project and three external economic factors. © The Author(s) 2025.

Instituciones: Pontificia Universidad Católica del Perú