0
  • DE
  • EN
  • FR
  • Base de données et galerie internationale d'ouvrages d'art et du génie civil

Publicité

Data Augmentation Approaches for Estimating Curtain Wall Construction Duration in High-Rise Buildings

Auteur(s): ORCID
ORCID

ORCID
Médium: article de revue
Langue(s): anglais
Publié dans: Buildings, , n. 4, v. 15
Page(s): 583
DOI: 10.3390/buildings15040583
Abstrait:

Reliable project management during planning stages of a building project is a meticulous process typically requiring sufficient precedencies. Typical construction duration estimation is based on previous cases of similar projects used to validate construction duration proposals from contractors, plan overall project duration, and set a standard for project success or failure. In cases of high-rise buildings exceeding 200 m, insufficient data commonly arise from the rarity of such projects, leading to a rough estimation of construction duration. Therefore, in this study, oversampling and data augmentation techniques derived from engineering principles, such as parametric optimization and data imbalance problems, are explored for curtain wall construction for high-rise buildings. The study was conducted in two phases. First, oversampling and data augmentation techniques, including Latin Hypercube, optimal Latin Hypercube, simple Monte Carlo, descriptive Monte Carlo, Sobol Monte Carlo, synthetic minority oversampling technique (SMOTE), and SMOTE–Tomek, were applied to 15 raw datasets collected from previous projects. The dataset was split into 8:2 for training and testing, where the mentioned techniques were applied to generate 500 virtual samples from the training data. Second, support vector regression was applied to forecast construction duration, where statistical performance criteria were applied for evaluation. The results showed that SMOTE and SMOTE–Tomek best represented the original dataset based on box plot analysis showcasing data distribution. Moreover, according to statistical performance criteria, it was found that the oversampling techniques improved the prediction performance, where Pearson correlation for linear, polynomial, and RBF increased by 0.611%, 4.232%, and 0.594%, respectively, for the best-performing sampling method. Finally, for the prediction models, probabilistic oversampling methods outperformed other methods according to the statistical performance criteria.

Copyright: © 2025 by the authors; licensee MDPI, Basel, Switzerland.
License:

Cette oeuvre a été publiée sous la license Creative Commons Attribution 4.0 (CC-BY 4.0). Il est autorisé de partager et adapter l'oeuvre tant que l'auteur est crédité et la license est indiquée (avec le lien ci-dessus). Vous devez aussi indiquer si des changements on été fait vis-à-vis de l'original.

  • Informations
    sur cette fiche
  • Reference-ID
    10820605
  • Publié(e) le:
    11.03.2025
  • Modifié(e) le:
    11.03.2025
 
Structurae coopère avec
International Association for Bridge and Structural Engineering (IABSE)
e-mosty Magazine
e-BrIM Magazine