Boosting-Based Ensemble Machine Learning Models for Predicting Unconfined Compressive Strength of Geopolymer Stabilized Clayey Soil
Scientists have developed new boosting-based ensemble machine learning models to accurately predict the unconfined compressive strength (UCS) of geopolymer stabilized clayey soil. This groundbreaking research, published in Scientific Reports, utilizes gradient boosting (GB) and adaptive boosting (AdaBoost) algorithms to improve the accuracy and reliability of UCS predictions.
The study focuses on geopolymer technology, which has gained significant attention in recent years for its potential to enhance the geotechnical properties of cohesive soils. Geopolymers are inorganic substances produced by synthesizing aluminosilicate compounds and are used as stabilizers to improve soil strength. This research explores the use of geopolymer, ground-granulated blast-furnace slag, fly ash, and sodium hydroxide solution to stabilize clayey soil samples.
To develop and validate their models, the researchers used a database of 270 clayey soil samples, randomly divided into training (80%) and testing (20%) sets. They assessed the accuracy and reliability of the GB and AdaBoost models using performance metrics such as the coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE), and mean squared error (MSE).
The results demonstrated that both GB and AdaBoost models outperformed other commonly used models, including artificial neural networks, random forest, extreme gradient boosting, multivariable regression, and multi-gen genetic programming. The GB and AdaBoost models achieved high values of R2 (0.980 and 0.975), low MAE (0.585 and 0.655), and low RMSE (0.969 and 1.088) for the testing dataset.
Furthermore, a sensitivity analysis revealed that the content of ground-granulated blast-furnace slag had the most significant impact on the UCS of geopolymer stabilized clayey soil, highlighting its importance as a key parameter in soil stabilization.
The study’s findings have significant implications for the field of geotechnical engineering. By accurately predicting the UCS of geopolymer stabilized clayey soil, engineers and researchers can make informed decisions regarding the design and construction of foundations, highways, dams, and other infrastructure projects. This research also highlights the potential of boosting-based ensemble machine learning models in geotechnical applications, offering a more efficient and reliable approach compared to traditional methods.
Geopolymer technology has emerged as an environmentally sustainable alternative to standard Portland cement due to its composition and production process. By utilizing industrial waste materials such as ground-granulated blast-furnace slag, metakaolin, and fly ash, the geopolymer industry aims to develop sustainable products with diverse applications in construction.
The study’s comprehensive literature review reveals that gradient boosting and adaptive boosting algorithms have not been extensively utilized to predict the UCS of geopolymer stabilized clayey soil. This research fills that gap by evaluating the potential of boosting-based machine learning models in accurately predicting the UCS, while considering the models’ applicability in civil engineering projects.
In summary, the development and validation of boosting-based ensemble machine learning models for predicting the UCS of geopolymer stabilized clayey soil represent a significant advancement in geotechnical engineering. This research contributes to the growing body of knowledge on geopolymer technology, highlighting its potential for improving soil properties and its wide range of applications in various construction projects.
