Strength prediction of concrete-filled steel tubular columns using Categorical Gradient Boosting algorithm

Due to complexities from the interaction between steel tube and concrete filling of concrete-filled tubular (CFST) columns, their strengths are very complicated, which is a highly nonlinear relation with material geometry. Categorical gradient Boosting (CatBoost), advanced boosting machine, presented solve problems. A total 3103 tests, divided in four datasets, trained tested learners determine ultimate axial strength as output variable while materials (concrete steel) geometry (e.g., diameters/width/heights, thickness, effective length, eccentricities) input ones. The comparison present results 10-fold cross validation those code predictions (AISC 360-16, Eurocode 4 AS/NZS 2327) previous study shows high prediction accuracy terms coefficient determination (R2), lowest value (R2 = 0.964) for Dataset 2 highest one 0.996) 1. While three codes beyond limit slenderness less conservative than within it, CatBoost provides nearly similar experiment mean values unity without any limits. This algorithm can be used predict an accurate CFST columns.