A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Compressive strength prediction of PET fiber-reinforced concrete using Dolphin echolocation optimized decision tree-based machine learning algorithms
This research presents a comprehensive study on predicting the compressive strength (CS) of PET-fiber-reinforced concrete (PFRC) using three decision tree-based machine learning models: Decision Tree (DT), Random Forest (RF), and Gradient Boosting Machine (GBM) regressors. To enhance the predictive capabilities of these models, the hyperparameters were optimized using the novel metaheuristic Dolphin Echolocation Optimization (DEO) technique. The input features considered for the models include the Binder content, W/B ratio, coarse and fine aggregate content, and PET fiber volume fraction. The target variable is the compressive strength of the concrete samples. Extensive experimentation was used to analyze and compare the effectiveness of each model. The results demonstrate that the DEO-tuned Random Forest outperformed its other counterparts, achieving improved accuracy in predicting the CS of PFRC. SHAP (Shapley Additive Explanations) and Sobol sensitivity analysis were conducted to explore the sensitivity of the input features toward compressive strength prediction. The Sobol sensitivity analysis assessed the significance of the input features and their interactions, whereas the SHAP values revealed the specific effects of each feature on the output of the model. The findings from the sensitivity analyses identified the Binder content, fiber volume fraction, and W/B as the most influential factors in determining the compressive strength.
Compressive strength prediction of PET fiber-reinforced concrete using Dolphin echolocation optimized decision tree-based machine learning algorithms
This research presents a comprehensive study on predicting the compressive strength (CS) of PET-fiber-reinforced concrete (PFRC) using three decision tree-based machine learning models: Decision Tree (DT), Random Forest (RF), and Gradient Boosting Machine (GBM) regressors. To enhance the predictive capabilities of these models, the hyperparameters were optimized using the novel metaheuristic Dolphin Echolocation Optimization (DEO) technique. The input features considered for the models include the Binder content, W/B ratio, coarse and fine aggregate content, and PET fiber volume fraction. The target variable is the compressive strength of the concrete samples. Extensive experimentation was used to analyze and compare the effectiveness of each model. The results demonstrate that the DEO-tuned Random Forest outperformed its other counterparts, achieving improved accuracy in predicting the CS of PFRC. SHAP (Shapley Additive Explanations) and Sobol sensitivity analysis were conducted to explore the sensitivity of the input features toward compressive strength prediction. The Sobol sensitivity analysis assessed the significance of the input features and their interactions, whereas the SHAP values revealed the specific effects of each feature on the output of the model. The findings from the sensitivity analyses identified the Binder content, fiber volume fraction, and W/B as the most influential factors in determining the compressive strength.
Compressive strength prediction of PET fiber-reinforced concrete using Dolphin echolocation optimized decision tree-based machine learning algorithms
Asian J Civ Eng
Parhi, Suraj Kumar (author) / Patro, Sanjaya Kumar (author)
Asian Journal of Civil Engineering ; 25 ; 977-996
2024-01-01
20 pages
Article (Journal)
Electronic Resource
English