摘要
本文研究了混凝土抗压强度与龄期和混凝土成分之间的关系,使用最小二乘法、最优子集选择、岭回归和Lasso回归建立了5个回归模型对混凝土抗压强度进行建模。通过建立训练集与测试集,计算测试集上的均方误差来评价模型的预测效果。结果显示,用交叉验证做最优子集选择拟合出的模型的均方误差是最小的,其次加权最小二乘回归的效果也比较好,它可以有较好的拟合优度,它的预测误差和其他几个模型的差距也不是很大。同时,模型也还有进一步改进的空间,可以考虑各特征之间的特性来构造出最合适的模型。
In this paper, the relationship between concrete compressive strength and age and concrete com-position was investigated and five regression models were developed to model concrete compres-sive strength using least squares, optimal subset selection, ridge regression and Lasso regression. The prediction effectiveness of the models was evaluated by establishing training and test sets and calculating the mean square error on the test set. The results show that the mean square error of the model fitted by using cross validation as the optimal subset selection is the smallest, followed by the weighted least squares regression, which can have a better goodness of fit, and the difference between its prediction error and several other models is not very large. At the same time, there is still room for further improvement of the model, and the characteristics between the features can be considered to construct the most suitable model.
出处
《应用数学进展》
2023年第9期3771-3784,共14页
Advances in Applied Mathematics