Estimation of Uniaxial Compressive Strength, Los Angeles Abrasion Loss and Frost Loss in Basalts with Flexible Calculation Methods

This study examines the estimation of Uniaxial Compressive Strength (UCS), Los Angeles abrasion loss (LA), and Na2SO4 Freeze Loss (FL) in basalts using flexible computational methods. The analysis was based on 67 experimental sets from Yozgat (Turkey) basalts, where properties such as dried unit weight, porosity, water absorption, and point load strength index were measured. Advanced regression models, including Random Forest, Multi-Layer Perceptron (MLP), Support Vector Regression (SVR), Linear Regression, and ExtraTree, were employed to predict key mechanical properties. Among these models, the ExtraTree algorithm achieved the highest performance, with R2 values of 1.000 for UCS, 0.999 for LA, and 1.000 for FL, and Variance Accounted For (VAF) scores of 99.980, 99.882, and 99.965%, respectively. These results highlight the model’s ability to accurately capture the complex, non-linear relationships between the input parameters and the mechanical properties of basalts. The study demonstrates that low-cost rapid tests, such as unit weight, porosity, water absorption, and point load strength index, can effectively predict critical basalt properties. This approach offers significant labour savings and improves efficiency in engineering applications. The strong performance of the ExtraTree and Random Forest models suggests their practical utility in a range of construction and geological engineering projects. Future research should focus on expanding the dataset to encompass a broader range of conditions and exploring additional features or ensemble methods to further enhance prediction accuracy. This study offers valuable insights into the use of machine learning techniques in geotechnical engineering, supporting more efficient, reliable, and cost-effective approaches to assessing the mechanical properties of basalts.