如何准确对岩体分级是深埋地下工程的一个难题.针对单一岩体评价模型存在误判的问题,提出了 Random Over Sampler(ROS)改进机器学习算法集成权重岩体质量评价方法.采用ROS算法对类别非均衡样本进行采样处理,经SVM、KNN、NB、DT、RBF、RF、LDA、LightGBM、XGBoost和GradientBoosting计算得到初步分级结果.将每一个分类器作为一个指标,确定每一个分类器的权重,建立ROS与机器学习算法的集成权重岩体质量评价模型,得到综合判别结果,大大降低了单一模型误判率.基于ROS改进机器学习算法集成权重提高了岩体质量评价模型的准确率,为岩体质量评价提供一种新的方法.
Integration of weighted rock mass quality evaluation method based on improved machine learning algorithm based and ROS
Accurate Rock mass classification is a significant challenge in deep underground engineering.To address the issue of misjudgment in a single rock mass evaluation model,an integrated weighted rock mass quality evaluation method based on the Random Over Sampler(ROS)improved machine learning algorithm is proposed.The ROS algorithm is used to oversample unbalanced samples,and the preliminary classification results are obtained by SVM,KNN,NB,DT,RBF,RF,LDA,LightGBM,XGBoost and GradientBoosting.Each classifier is treated as an index to determine its weight,and an integrated weight rock mass quality evaluation model combining ROS and machine learning algorithm is established.This approach significantly reduces the misjudgment rate of individual models.The results demonstrate that the ROS-based improved machine learning algorithm increases the accuracy of rock mass quality evaluation models,providing a novel method for rock mass quality assessment.
NETL
NSTL
万方数据
