基于机器学习的盾构掘进地表沉降回归预测模型

针对地铁盾构掘进引起的地表沉降预测问题,经验和半经验公式预测差异性大,解析法预测过程复杂,数值模拟无法运用于实时预测。基于线性回归、岭回归、套索算法(least absolute shrinkage and selection operator, LASSO)、分裂决策树算法(classification and regression tree, CART)、随机森林算法、支持向量机(support vector machine, SVM)、XGBoost(extreme gradient boosting)、多层感知机(multi-layer perceptron, MLP)8种机器学习算法的拟合能力,文章提出盾构掘进沿线地表最大沉降值预测模型。采用均方误差(mean square error, MSE)、均方根误差(root mean square error, RMSE)、绝对均值误差(mean absolute error, MAE)、决定系数 $ R^{2} $进行模型评估。结果表明:支持向量回归(support vector regression, SVR)模型取得的效果最优,其次是回归MLP模型;在树模型中,随机森林的表现效果好于CART和XGBoost;线性模型中表现最好的是岭回归。

In view of the prediction of ground settlement caused by subway shield tunneling, there is a big difference between empirical and semi-empirical formula prediction, and the prediction process of the analytic method is complicated, so the numerical simulation cannot be applied to real-time prediction. Based on the fitting ability of eight machine learning algorithms, such as linear regression, ridge regression, least absolute shrinkage and selection operator (LASSO), classification and regression tree (CART), random forest, support vector machine (SVM), extreme gradient boosting (XGBoost) and multi-layer perceptron (MLP), a prediction model of the maximum ground settlement along the shield tunneling line is proposed. The mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination ( $ R^{2} $) were used to evaluate the model. The results show that support vector regression (SVR) model has the best effect, followed by regression MLP model. In the tree model, the performance of random forest is better than that of CART and XGBoost. Ridge regression performs best in the linear models.