Prediction of Unconfined Compressive Strength in Cement-Treated Soil: A Machine Learning Approach

This study explores machine learning approaches for predicting unconfined compressive strength (UCS) of cement-treated clayey silt from northwestern Iași, Romania. Laboratory testing generated 185 UCS measurements examining the influence of cement content, curing period, and compaction rate. Exploratory data analysis revealed strong correlations between UCS and cement content and moderate correlations with curing period, while compaction rate showed minimal impact. Multiple regression algorithms were evaluated using a nested cross-validation framework with Bayesian hyperparameter optimization. Random Forest demonstrated superior performance when trained on the complete dataset. Feature importance analysis confirmed cement content and curing period as the dominant factors influencing strength development. The model enables geotechnical engineers to predict stabilized soil strength without extensive laboratory testing, supporting efficient design of ground improvement works. The methodology demonstrates the potential of data-driven approaches in geotechnical engineering applications, with both original data and the trained model made publicly available to support further research and practical implementation.