In EDS ansehen

Anticipating Properties of Ultra-High-Performance Concrete through the K-Nearest Neighbor Algorithm

Gespeichert in:
Bibliographische Detailangaben
Titel: Anticipating Properties of Ultra-High-Performance Concrete through the K-Nearest Neighbor Algorithm
Autoren: Yonghoon Won
Quelle: Journal of Artificial Intelligence and System Modelling, Vol 03, Iss 03, Pp 15-33 (2025)
Verlagsinformationen: Bilijipub publisher, 2025.
Publikationsjahr: 2025
Bestand: LCC:Electronic computers. Computer science
Schlagwörter: ultra-high-performance concrete, machine learning, k-nearest neighbor-based, giant trevally optimizer, population-based vortex search algorithm, Electronic computers. Computer science, QA75.5-76.95
Beschreibung: The high magnitude of compressive strength (CS) in ultra-high-performance concrete (UHPC) is strongly related to the specific characteristics, proportions, and composition of the constituent materials. In decoding such a complex relationship, There is a pressing necessity to employ Machine Learning (ML) and artificial intelligence (AI) methodologies. The K-nearest neighbor (KNN) method has proved superior in developing predictive models that perfectly harmonize with experimental datasets. It is worth mentioning its remarkable accuracy: it accurately reflects experimental results and shows the efficacy of KNN in predicting the behavior of UHPC with regard to input parameters. Two state-of-the-art metaheuristic approaches, namely Giant Trevally Optimizer (GTO) and Population-based Vortex Search Algorithm (PBVSA), have been combined with KNN in this study for enhanced predictive accuracy. The combination of these algorithms yields three hybrid models, namely KNN+GTO (KNGT), KNN+PBVSA (KNVS), and KNN. Considering its far-superior R2 values, the KNGT model is obviously the outrunner. Being successful during the training phase itself by giving an R2 value of 0.995 with an ideal RMSE of 2.207, the metrics of the KNGT model mark it out from different models developed in this research for unmatched predictive and generalization abilities. The present study has essentially coupled advanced ML with the KNN algorithm, strategic integration of GTO and PBVSA techniques, and experiments based on empiricism. Thus, additional predictive strength has been conferred on UHPC-CS projections. This synthesis of advanced methodologies strengthens the understanding of UHPC behavior and provides a strong method for precisely anticipating the CS characteristics in this context.
Publikationsart: article
Dateibeschreibung: electronic resource
Sprache: English
ISSN: 3041-850X
Relation: https://jaism.bilijipub.com/article_230939_9da4fee65acae5a0b493196ad3f73fc8.pdf; https://doaj.org/toc/3041-850X
DOI: 10.22034/jaism.2025.527639.1135
Zugangs-URL: https://doaj.org/article/ac83ac0a8b514aab81a1fd9d7fc7887b
Dokumentencode: edsdoj.83ac0a8b514aab81a1fd9d7fc7887b
Datenbank: Directory of Open Access Journals
Beschreibung
Abstract:The high magnitude of compressive strength (CS) in ultra-high-performance concrete (UHPC) is strongly related to the specific characteristics, proportions, and composition of the constituent materials. In decoding such a complex relationship, There is a pressing necessity to employ Machine Learning (ML) and artificial intelligence (AI) methodologies. The K-nearest neighbor (KNN) method has proved superior in developing predictive models that perfectly harmonize with experimental datasets. It is worth mentioning its remarkable accuracy: it accurately reflects experimental results and shows the efficacy of KNN in predicting the behavior of UHPC with regard to input parameters. Two state-of-the-art metaheuristic approaches, namely Giant Trevally Optimizer (GTO) and Population-based Vortex Search Algorithm (PBVSA), have been combined with KNN in this study for enhanced predictive accuracy. The combination of these algorithms yields three hybrid models, namely KNN+GTO (KNGT), KNN+PBVSA (KNVS), and KNN. Considering its far-superior R2 values, the KNGT model is obviously the outrunner. Being successful during the training phase itself by giving an R2 value of 0.995 with an ideal RMSE of 2.207, the metrics of the KNGT model mark it out from different models developed in this research for unmatched predictive and generalization abilities. The present study has essentially coupled advanced ML with the KNN algorithm, strategic integration of GTO and PBVSA techniques, and experiments based on empiricism. Thus, additional predictive strength has been conferred on UHPC-CS projections. This synthesis of advanced methodologies strengthens the understanding of UHPC behavior and provides a strong method for precisely anticipating the CS characteristics in this context.
ISSN:3041850X
DOI:10.22034/jaism.2025.527639.1135