A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Predictive modelling of surface chloride concentration in marine concrete structures: a comparative analysis of machine learning approaches
This study focuses on predicting surface chloride concentration (Cs) in marine concrete structures using machine learning (ML) models. The dataset includes input features related to composition and environmental conditions, along with corresponding Cs values. Objectives include evaluating ML model performance, investigating the impact of outliers, analyzing feature importance, and proposing an adaptive ensemble model. ML models used are linear regression, support vector machine regression, decision tree regression, random forest regression, gradient boosting regression, XGBoost regression, and multi-layer perceptron regression. The dataset is pre-processed with normalization, outlier handling, and one-hot encoding for categorical features. Models are trained with tenfold cross-validation. Outliers handling has varying effects on model performance. Linear regression, decision tree, and XGBoost models show decreased performance, while support vector machine regression improves performance. Random Forest, gradient boosting, multi-layer perceptron, and the adaptive ensemble model are relatively unaffected. Feature importance analysis reveals varying significance of input features across models. The Adaptive Ensemble model combines predictions from base models using weighted averaging. Weights are assigned based on performance. The proposed ensemble model achieves improved predictive accuracy (around 95%) compared to individual models (best individual model performance around 81%).
Predictive modelling of surface chloride concentration in marine concrete structures: a comparative analysis of machine learning approaches
This study focuses on predicting surface chloride concentration (Cs) in marine concrete structures using machine learning (ML) models. The dataset includes input features related to composition and environmental conditions, along with corresponding Cs values. Objectives include evaluating ML model performance, investigating the impact of outliers, analyzing feature importance, and proposing an adaptive ensemble model. ML models used are linear regression, support vector machine regression, decision tree regression, random forest regression, gradient boosting regression, XGBoost regression, and multi-layer perceptron regression. The dataset is pre-processed with normalization, outlier handling, and one-hot encoding for categorical features. Models are trained with tenfold cross-validation. Outliers handling has varying effects on model performance. Linear regression, decision tree, and XGBoost models show decreased performance, while support vector machine regression improves performance. Random Forest, gradient boosting, multi-layer perceptron, and the adaptive ensemble model are relatively unaffected. Feature importance analysis reveals varying significance of input features across models. The Adaptive Ensemble model combines predictions from base models using weighted averaging. Weights are assigned based on performance. The proposed ensemble model achieves improved predictive accuracy (around 95%) compared to individual models (best individual model performance around 81%).
Predictive modelling of surface chloride concentration in marine concrete structures: a comparative analysis of machine learning approaches
Asian J Civ Eng
Tipu, Rupesh Kumar (author) / Batra, Vandna (author) / Suman (author) / Panchal, V. R. (author) / Pandya, K. S. (author)
Asian Journal of Civil Engineering ; 25 ; 1443-1465
2024-02-01
23 pages
Article (Journal)
Electronic Resource
English
Modelling chloride ingress in marine environment concrete
| Tema Archive | 2009
Approaches for Modelling the Residual Service Life of Marine Concrete Structures
| BASE | 2014
Approaches for Modelling the Residual Service Life of Marine Concrete Structures
| Tema Archive | 2014