Evaluation of Unconfined Compressive Strength of Cement using Ensemble Learning Regression

Machine Learning (ML) technologies have emerged for predicting mechanical properties of CPB (Cemented Paste Backfill). But due to a lack of more robust methodologies, their applications have not yet reached their full potential. The state-of-the-art collective learning methodology is used in this paper for improving the evaluation of cement's unconfined compressive strength (UCS). Individual ML techniques were chosen from regression approaches such as Ridge Regression, Linear regression, Support Vector Regression, and Lasso Regression. The Decision Tree Regressor and Random Forest Regressor ensemble learning frameworks were also used. Cross validation of 10-folds was utilized as the method of confirmation and the execution was assessed at 95% confidence interval. Alongside, hyper-parameters tuning was conducted. The finest result is obtained by taking a ratio of 70:30 for training/test set. The approach provided on this observation expands latest efforts for UCS foresight of cement and may extensively boost up the cement design.