Robust Anomaly Recognition in Hydraulic Structural Safety Monitoring: A Methodology Based on Deconfounding Boosted Regression Trees

Algorithm 1

Training of a two-stage boosted regression trees model.

Input: Train data collection , loss function L, max iteration number M, and learning rate λ

Output: Second-stage regression function and each first stage regression function 1//Initialize each first stage regression function2//Initialize second-stage regression function3whiledo4  whiledo5    //Get residual’s gradient of 6   7   8   //Train a new CART tree to update regression function of 9  end10   //Get residual’s gradient of 11  12  13  //Train a new CART tree to update regression function of 14end15Return and