A comparative assessment of tree-based predictive models to estimate geopolymer concrete compressive strength

Fly ash-based geopolymer concrete (FA-GPC) is a material that might be utilized to build a more sustainable construction industry; therefore, this paper aims to develop a novel approach for its compressive strength (CS) prediction. To achieve this goal, three tree-based machine learning methods, nam...

Full description

Saved in:
Bibliographic Details
Published in:Neural computing & applications 2023-03, Vol.35 (9), p.6569-6588
Main Authors: Nguyen, May Huu, Mai, Hai-Van Thi, Trinh, Son Hoang, Ly, Hai-Bang
Format: Article
Language:English
Subjects:
Algorithms
Artificial Intelligence
Building materials industry
Cement
Climate change
Coal-fired power plants
Compressive strength
Computational Biology/Bioinformatics
Computational Science and Engineering
Computer Science
Concrete
Concrete properties
Construction industry
Curing
Data Mining and Knowledge Discovery
Decision trees
Emissions
Fly ash
Geopolymers
Greenhouse gases
Image Processing and Computer Vision
Industrial plant emissions
Machine learning
Original Article
Prediction models
Probability and Statistics in Computer Science
Sensitivity analysis
Silica
Silicon dioxide
Zeolites
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Fly ash-based geopolymer concrete (FA-GPC) is a material that might be utilized to build a more sustainable construction industry; therefore, this paper aims to develop a novel approach for its compressive strength (CS) prediction. To achieve this goal, three tree-based machine learning methods, namely, Radom Forest (RF), Decision Tree (DT), and Extreme Gradient Boosting (XGBoost), were developed based on experimental databases considering fourteen key factors of FA-GPC. The results indicated that XGBoost outperformed other methods, as proved by excellent prediction metrics. The sensitivity analysis of XGBoost models was then discussed utilizing feature importance, mean Shapley additive explanations (SHAP), and Beeswarm-SHAP values. The findings indicated that FA content had the most crucial impact on the CS, followed by SiO 2 and NaOH contents, among the other variables examined. Finally, the SHAP dependence plots technique was utilized to quantitatively discuss feature interactions and contributions to the CS of FA-GPC.
ISSN:0941-0643
1433-3058
DOI:10.1007/s00521-022-08042-2