Prediction model for concrete compressive strength based on XGBoost hyperparametric optimization with Optuna
This study built a robust nonlinear forecasting model leveraging the XGBoost gradient boosting algorithm,enhanced through hyperparameter optimization via Optuna,to accurately estimate concrete compressive strength,thereby improving engineering design level and construction quality.A comparative analysis was conducted for the predictive accuracy of XGBoost against other algorithms—LightGBM,NGBoost,and CatBoost—utilizing a dataset comprising 1110 unique concrete test samples from existing literature.The evaluation demonstrated that XGBoost outperformed these alternatives in terms of accuracy for both training and testing datasets.Subsequently,Optuna was applied to fine-tune the hyperparameters of the best-performing XGBoost model.The results reveal that the hyperparameter optimization significantly enhances the predictive precision of the XGBoost model.In testing datasets,the optimized XGBoost model demonstrated strong generalization capabilities,confirming its effectiveness in predicting concrete strength and serving as a valuable reference for future engineering practices.
万方数据
维普
