A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
A Novel Multiple Linear Regression Approach for Predicting the Unconfined Compressive Strength of Soil
https://orcid.org/0000-0003-2462-980X
This paper proposes a precise and general multiple linear regression (MLR) model to predict the unconfined compressive strength (UCS) of various soil types. The study used a wide range of data sets, including 952 data points considering 39 soil types with varying grain sizes. The model inputs were soil physical properties, grain size, age, mixture proportion, and chemical composition of binder materials. An innovative and novel approach was developed to enhance the accuracy of the MLR model, a randomized exploratory algorithm. The model demonstrated significant accuracy with a 0.921 R2 in the testing data set. The Bayesian model averaging (BMA) method was employed for feature reduction, focusing on important variables. Alternative models were also developed based on the significant variables highlighted by the BMA approach, all showing high accuracy in predicting the UCS. The proposed models demonstrated superiority over traditional approaches based on the data set size and statistical metrics. The paper provides instances of predicting soil UCS and determining the mix design corresponding to the target UCS.
A Novel Multiple Linear Regression Approach for Predicting the Unconfined Compressive Strength of Soil
https://orcid.org/0000-0003-2462-980X
This paper proposes a precise and general multiple linear regression (MLR) model to predict the unconfined compressive strength (UCS) of various soil types. The study used a wide range of data sets, including 952 data points considering 39 soil types with varying grain sizes. The model inputs were soil physical properties, grain size, age, mixture proportion, and chemical composition of binder materials. An innovative and novel approach was developed to enhance the accuracy of the MLR model, a randomized exploratory algorithm. The model demonstrated significant accuracy with a 0.921 R2 in the testing data set. The Bayesian model averaging (BMA) method was employed for feature reduction, focusing on important variables. Alternative models were also developed based on the significant variables highlighted by the BMA approach, all showing high accuracy in predicting the UCS. The proposed models demonstrated superiority over traditional approaches based on the data set size and statistical metrics. The paper provides instances of predicting soil UCS and determining the mix design corresponding to the target UCS.
A Novel Multiple Linear Regression Approach for Predicting the Unconfined Compressive Strength of Soil
https://orcid.org/0000-0003-2462-980X
This paper proposes a precise and general multiple linear regression (MLR) model to predict the unconfined compressive strength (UCS) of various soil types. The study used a wide range of data sets, including 952 data points considering 39 soil types with varying grain sizes. The model inputs were soil physical properties, grain size, age, mixture proportion, and chemical composition of binder materials. An innovative and novel approach was developed to enhance the accuracy of the MLR model, a randomized exploratory algorithm. The model demonstrated significant accuracy with a 0.921 R2 in the testing data set. The Bayesian model averaging (BMA) method was employed for feature reduction, focusing on important variables. Alternative models were also developed based on the significant variables highlighted by the BMA approach, all showing high accuracy in predicting the UCS. The proposed models demonstrated superiority over traditional approaches based on the data set size and statistical metrics. The paper provides instances of predicting soil UCS and determining the mix design corresponding to the target UCS.
A Novel Multiple Linear Regression Approach for Predicting the Unconfined Compressive Strength of Soil
Int. J. Geomech.
Mahmoudi, Mohammadreza (author) / Toufigh, Vahab (author) / Ghaemian, Mohsen (author)
2024-08-01
Article (Journal)
Electronic Resource
English
| British Library Online Contents | 2012