Research on Long-term Deflection Prediction of Reinforced Concrete Beams Based on XGBoost
YUE Xinxin
CHANG Shan
DU Yujie
MA Lu
DAN Wenjiao
Abstract：To accurately predict the long-term deflection of reinforced concrete(RC)beams,long-term deflection test data of RC beams were collected,and a prediction model was established based on eXtreme Gradient Boosting(XGBoost).The accuracy of this model was tested and compared with prediction models established using Support Vector Regression(SVR)and Back Propagation Neural Network(BPNN)on the test dataset.The results show that the long-term deflection prediction model for RC beams based on XGBoost model can be effectively used for long-term deflection prediction,achieving coefficients of determination reaching 1.000 0 and 0.981 8 on the training set and testing set,respectively.Comparing with the models established using SVR and BPNN,the root mean square error of XGBoost model was reduced by 98.03％,15.93％for the training set,and by 99.47％and 85.97％for the testing set.Additionally,the mean absolute percentage error was reduced by 95.51％,11.81％for the training set,and by 96.40％and 30.83％for the testing set.Finally,a global sensitivity analysis of the long-term deflection of RC beams based on XGBoost model was conducted to rank the importance of influential parameters.The results demonstrate that the XGBoost model has the excellent performance.
Keywords：reinforced concrete beamXGBoostLong-term deformation deflection predictionmachine learningglobal sensitivity analysis
Publication Date：2024-07-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：8( 98-105 )