Regression-based models for the prediction of unconfined compressive strength of artificially structured soil

Unconfined compressive strength (UCS) of soil is a critical and important geotechnical property which is widely used as input parameters for the design and practice of various geoengineering projects. UCS controls the deformational behavior of soil by measuring its strength and load bearing capacity. The laboratory determination of UCS is tedious, expensive and being a time-consuming process. Therefore, the present study is aimed to establish empirical equations for UCS using simple and multiple linear regression methods. The accuracy of the developed equations are tested by employing coefficient of determination (R2), root mean square error (RMSE) and mean absolute percentage error (MAPE). It has been found that the developed equations are reliable and capable to predict UCS with acceptable degree of confidence. Among all the developed models, model-I consist of lime content, curing time, plastic limit, liquid limit, potential of hydrogen, primary ultrasonic wave velocity, optimum moisture content and maximum dry density as independent parameters shows highest prediction capacity with R2, RMSE and MAPE are 0.96, 25.89 and 16.59, respectively.