A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Utilization of Tree-Based Ensemble Models for Predicting the Shear Strength of Soil
https://orcid.org/http://orcid.org/0000-0001-8536-5305
One of the most significant factors associated with the geotechnical design method is the precise determination of the shear strength of the soil (SSS). Geotechnical experts must be able to accurately predict the SSS without requiring costly laboratory testing, which is a practical requirement. A dataset of 249 soil samples and six controlling variables (depth, clay proportion, loam proportion, sand proportion, plastic limit, and plastic index) were utilized to build an algorithm for predicting the SSS. To this end, five tree-breed models, (i) extra tree regressor (ETR), (ii) decision tree regressor (DTR), (iii) ridge regression (RR), (iv) linear regression (LR), and (v) Bayesian ridge regressor (BRR), were developed. The developed models were verified and assessed using six performance indices. After validation of ensemble models, it was found that, out of five, ETR produces good modeling outcomes, outperforming other models. The rank analysis of the models produced also confirmed this. To determine the importance of the different influencing variables used to predict the outcome, a sensitivity analysis was performed. The best model’s learning process was also carried out utilizing model-based learning curves, outlier distance detection plots, feature importance plots, prediction error plots, and residual plots. The feature significance plot revealed that sample depth is the strongest feature that contributes to SSS prediction in the best model.
Utilization of Tree-Based Ensemble Models for Predicting the Shear Strength of Soil
https://orcid.org/http://orcid.org/0000-0001-8536-5305
One of the most significant factors associated with the geotechnical design method is the precise determination of the shear strength of the soil (SSS). Geotechnical experts must be able to accurately predict the SSS without requiring costly laboratory testing, which is a practical requirement. A dataset of 249 soil samples and six controlling variables (depth, clay proportion, loam proportion, sand proportion, plastic limit, and plastic index) were utilized to build an algorithm for predicting the SSS. To this end, five tree-breed models, (i) extra tree regressor (ETR), (ii) decision tree regressor (DTR), (iii) ridge regression (RR), (iv) linear regression (LR), and (v) Bayesian ridge regressor (BRR), were developed. The developed models were verified and assessed using six performance indices. After validation of ensemble models, it was found that, out of five, ETR produces good modeling outcomes, outperforming other models. The rank analysis of the models produced also confirmed this. To determine the importance of the different influencing variables used to predict the outcome, a sensitivity analysis was performed. The best model’s learning process was also carried out utilizing model-based learning curves, outlier distance detection plots, feature importance plots, prediction error plots, and residual plots. The feature significance plot revealed that sample depth is the strongest feature that contributes to SSS prediction in the best model.
Utilization of Tree-Based Ensemble Models for Predicting the Shear Strength of Soil
https://orcid.org/http://orcid.org/0000-0001-8536-5305
One of the most significant factors associated with the geotechnical design method is the precise determination of the shear strength of the soil (SSS). Geotechnical experts must be able to accurately predict the SSS without requiring costly laboratory testing, which is a practical requirement. A dataset of 249 soil samples and six controlling variables (depth, clay proportion, loam proportion, sand proportion, plastic limit, and plastic index) were utilized to build an algorithm for predicting the SSS. To this end, five tree-breed models, (i) extra tree regressor (ETR), (ii) decision tree regressor (DTR), (iii) ridge regression (RR), (iv) linear regression (LR), and (v) Bayesian ridge regressor (BRR), were developed. The developed models were verified and assessed using six performance indices. After validation of ensemble models, it was found that, out of five, ETR produces good modeling outcomes, outperforming other models. The rank analysis of the models produced also confirmed this. To determine the importance of the different influencing variables used to predict the outcome, a sensitivity analysis was performed. The best model’s learning process was also carried out utilizing model-based learning curves, outlier distance detection plots, feature importance plots, prediction error plots, and residual plots. The feature significance plot revealed that sample depth is the strongest feature that contributes to SSS prediction in the best model.
Utilization of Tree-Based Ensemble Models for Predicting the Shear Strength of Soil
Transp. Infrastruct. Geotech.
Rabbani, Ahsan (author) / Muslih, Jan Afzal (author) / Saxena, Mukul (author) / Patil, Santosh Kalyanrao (author) / Mulay, Bharat Nandkumar (author) / Tiwari, Mohit (author) / Usha, A (author) / Kumari, Sunita (author) / Samui, Pijush (author)
Transportation Infrastructure Geotechnology ; 11 ; 2382-2405
2024-08-01
24 pages
Article (Journal)
Electronic Resource
English
Utilization of Tree-Based Ensemble Models for Predicting the Shear Strength of Soil
| Springer Verlag | 2024
Implementing ensemble learning models for the prediction of shear strength of soil
| Springer Verlag | 2023
| NTIS | 1973