A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Comparative Analysis of Automated Machine Learning and Optimized Conventional Machine Learning for Concrete’s Uniaxial Compressive Strength Prediction
The uniaxial compressive strength (UCS) is a crucial mechanical property influenced by factors such as concrete constituents and curing days. Concrete’s UCS poses significant challenges for accurate estimation. Traditional methods are time-intensive, expensive, and may struggle to account for the impact of various interacting factors. This study pioneers the application of automated machine learning (AutoML) and conventional ML techniques to unravel the intricate relationships between the UCS and six factors. A robust dataset comprising 844 experimental results was used to train and evaluate the models. The input parameters for the models were: the curing days, amount of plasticizer, and quantity of cement, fine and coarse aggregates (CAs). Among the models assessed, the AutoGluon model stands out for its superior prediction accuracy and result interpretability. AutoGluon showed exceptional performance when the predictions were compared with experimental data. This model yielded the lowest root mean square error (RMSE) of 1.0830 MPa and the highest coefficient of determination (R2) of 0.9493. Analysis of feature importance indicates that curing days of concrete is the most influential parameter for this prediction task. The study demonstrates that the AutoGluon model reliably and robustly estimates concrete’s UCS. Additionally, they are more efficient to train than conventional ML models, eliminating the need for the laborious and time-consuming process of hyperparameter tuning. Specifically, in assessing the UCS of concrete AutoGluon had 0.64%–1.82% and 0.07%–0.2% more superior RMSE and R2 than the conventional ML models.
Comparative Analysis of Automated Machine Learning and Optimized Conventional Machine Learning for Concrete’s Uniaxial Compressive Strength Prediction
The uniaxial compressive strength (UCS) is a crucial mechanical property influenced by factors such as concrete constituents and curing days. Concrete’s UCS poses significant challenges for accurate estimation. Traditional methods are time-intensive, expensive, and may struggle to account for the impact of various interacting factors. This study pioneers the application of automated machine learning (AutoML) and conventional ML techniques to unravel the intricate relationships between the UCS and six factors. A robust dataset comprising 844 experimental results was used to train and evaluate the models. The input parameters for the models were: the curing days, amount of plasticizer, and quantity of cement, fine and coarse aggregates (CAs). Among the models assessed, the AutoGluon model stands out for its superior prediction accuracy and result interpretability. AutoGluon showed exceptional performance when the predictions were compared with experimental data. This model yielded the lowest root mean square error (RMSE) of 1.0830 MPa and the highest coefficient of determination (R2) of 0.9493. Analysis of feature importance indicates that curing days of concrete is the most influential parameter for this prediction task. The study demonstrates that the AutoGluon model reliably and robustly estimates concrete’s UCS. Additionally, they are more efficient to train than conventional ML models, eliminating the need for the laborious and time-consuming process of hyperparameter tuning. Specifically, in assessing the UCS of concrete AutoGluon had 0.64%–1.82% and 0.07%–0.2% more superior RMSE and R2 than the conventional ML models.
Comparative Analysis of Automated Machine Learning and Optimized Conventional Machine Learning for Concrete’s Uniaxial Compressive Strength Prediction
Chukwuemeka Daniel (author)
2024
Article (Journal)
Electronic Resource
Unknown
Metadata by DOAJ is licensed under CC BY-SA 1.0
Prediction of Uniaxial Compressive Strength of Rock Using Machine Learning
| Springer Verlag | 2022
Estimation of Intact Rock Uniaxial Compressive Strength Using Advanced Machine Learning
| Springer Verlag | 2024
Estimation of Intact Rock Uniaxial Compressive Strength Using Advanced Machine Learning
| Springer Verlag | 2024