<![CDATA[Accurate prediction of concrete compressive strength is critical for quality control, structural evaluation, and mix optimization. However, the connection between mix composition, curing age, and compressive strength is often nonlinear and heterogeneous. This study assesses four prediction models, including Native Gaussian Process Regression, Gaussian Process Regression with Conformal Intervals, Support Vector Regression, and K-Nearest Neighbors, to estimate concrete compressive strength at various curing ages. Predictors are produced from a combination of composition, curing age, and material-important properties. Model performance is evaluated using Leave-One-Age-Out and repeated stratified K-fold by Age, with root mean square error, mean absolute error, coefficient of determination, coverage, and average interval length. The results demonstrate that Native Gaussian Process Regression gives the best steady overall performance in both assessment methodologies. At the same time, Gaussian Process Regression with Conformal Intervals tends to provide higher coverage with the consequence of wider intervals. In general, Native Gaussian Process Regression provides the best balance among point prediction accuracy, interval coverage, and interval efficiency for heterogeneous concrete compressive strength data by curing age.]]>
Copyrights © 2026
