<p>Elastic modulus is a vital parameter of geotechnical design and cannot be easily measured directly as it is costly and time-consuming. This study developed a laboratory database of 147 sedimentary rocks specimens and tested four tree-ensemble models to estimate elastic modulus. It used eight physical parameters such as porosity, dry density, saturated density, water absorption and ultrasonic velocities both in dry and saturated conditions as the predictors. Box-plot and Z-score have been used to remove outliers. A severe multicollinearity was identified based on the variance inflation factor analysis and dissolved using principal component analysis (PCA). Five-fold cross-validation was used to train the models and the metrics used to evaluate the model included, coefficient of determination, root mean square error, mean absolute error, weighted mean absolute percentage error and bias. Category Gradient Boosting (CatBoost) was the best in terms of performance and the coefficient of determination was 0.889, the root mean square error was 1.06 GPa, and the mean absolute error was 0.75 GPa. Statistical analysis proved that there were significant differences in the performance of models. Findings have shown that the suggested framework can give precise and credible results on predicting the elastic modulus through non-destructive laboratory parameters.</p>