BAS-RF组合算法在大坝水平位移预测中的应用

Application of BAS-RF Combined Algorithm in Prediction of Dam Horizontal Displacement

针对大坝变形监测数据呈现非线性、影响因素复杂和预测性能不佳等问题,提出了天牛须搜索(Beetle Antennae Search,BAS)算法和随机森林(Random Forest,RF)算法相结合的大坝水平位移预测模型。利用BAS算法收敛速度快、计算量小的特性,对RF预测模型中的参数决策树个数和分裂属性个数进行深度优化,选取最佳参数构建模型。通过选取实训数据,与LSSVM,BP,RF模型进行大坝水平位移预测对比分析,结果表明,BAS-RF组合模型在大坝水平位移预测中具有较好的稳定性,变形预测精度也得到提高。

To solve the problems of non-linearity of dam deformation monitoring data,complex influencing factors,low prediction performance,etc.,a dam horizontal displacement prediction model combining Beetle Antennae Search(BAS)algorithm and random forest algorithm is proposed.Taking advantage of the fast convergence speed and low computational complexity of BAS algorithm,the number of parameter decision trees and split attributes in the random forest prediction model is deeply optimized,and the best parameters are selected to build the model.Through the selection of training data,the proposed model of dam horizontal displacement prediction is compared with the LSSVM,BP,and RF models.The results show that the BAS-RF combined model has better stability in dam horizontal displacement prediction,and the accuracy of the deformation prediction is also improved.