基于Xgboost优化的KELM滑坡预报模型研究  

Research on the Optimization KELM Landslide Prediction Modeling Based on Xgboost

作　　者：李璐 徐根祺 杨倩[1] 王艳娥 赵正健 LI Lu;XU Genqi;YANG Qian;WANG Yane;ZHAO Zhengjian(Xi'an Siyuan University of Science and Technology,Xi'an 710038,China;School of Mechanical and Electrical Engineering,Xi'an Transportation Engineering College,Xi'an 710030,China)

机构地区：[1]西安思源学院理工学院,西安710038 [2]西安交通工程学院机械与电气工程学院,西安710030

出　　处：《计算机测量与控制》2023年第4期225-231,共7页Computer Measurement &Control

基　　金：陕西省教育厅科研计划资助项目(2022JK0515);陕西省自然科学基础研究计划项目(2023-JC-YB-464)。

摘　　要：针对极限学习机对滑坡预测准确性低及在训练过程中模型不稳定的问题,引入RBF高斯核函数并使用极限梯度提升树算法Xgboost对KELM进行优化,建立了Xgboost优化后的Xgboost-KELM预测模型;首先采用高斯核RBF作为极限学习机的核函数,解决隐藏节点随机映射问题,增加模型稳定性及适用性;其次将清洗后的监测数据作为模型输入,并使用Xgboost寻优算法对核函数中的超参数进行优化,通过4组测试集进行Xgboost-KELM建模,依据均方误差迭代曲线得出最佳超参数;最后使用两组10%样本集验证模型评价指标及稳定性,实验结果AUC均值对比模型至少提高3个百分点,Precision、Accuracy及Recall至少高于对比模型1.7个百分点,同时Xgboost-KELM模型的方差及偏差都较小,证明该模型稳定性较好,实验结果说明Xgboost-KELM模型具有较好的预测效果,在滑坡灾害预测中有较好的预测能力。To solve the problems of low accuracy for an extreme learning machine(ELM)in landslide prediction,and the instability of the model in the training process,a radical basis function(RBF)Gaussian kernel function is introduced to optimize the Kernel extreme learning machine(KELM)by using a xtreme gradient boosting(Xgboost)algorithm,and the Xgboost KELM prediction model of the optimizated Xgboost is established;Firstly,as the kernel function of the limit learning machine,the Gaussian kernel RBF is used to solve the random mapping of the hidden nodes and increase the stability and applicability of the model;Secondly,the cleaned monitoring data is used as the model input,and the Xgboost optimization algorithm is used to optimize the super parameters in the kernel function,the Xgboost KELM modeling is conducted through four groups of test sets,and the best super parameters are obtained according to the iteration curve of the mean square error;Finally,two groups of 10%sample sets are used to verify the model evaluation indicators and stability.The experimental results show that compared with the GA and GC models,the AUC mean of the Xgboost-KELM model is increased by 3 percentage points,and the performance indexes of precision,accuracy and recall by at least 1.7 percentage points.At the same time,the variance and deviation of the Xgboost KELM model are small,which proves that the model is stable and has a good prediction ability in landslide disaster prediction.

关 键 词：高斯核RBF KELM Xgboost超参数 滑坡灾害 预报模型 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]