基于SMOTE算法的岩爆烈度等级预测模型研究

打开文本图片集

中图分类号：U45 文献标识码：A 文章编号：2097-5465（2025）03-0030-08

1.Hebei GEO University，Shijiazhuang O50031，China；2.Hebei Technology Innovation Center for Intelligent

Development and Control of Underground Built Environment，Shijiazhuang O5o031，China

Abstract：Inordertosolvetheproblemofdataimbalanceintherockburstdatabase，resultinginlowpredictionaccuracyof rockburst，fivemodelswereproposedbasedonthesyntheticminorityoversampling technique（SMOTE），including SMOTErandomforest，SMOTE-gradientboostingdecisiontree，SMOTE-supprt vector machine，SMOTE-BP neural network andSMOTEconvolutionalneuralnetwork.Inthispaper，sixindicatorswereselectedandtherockburstintensitygradewasdividedintofour grades，oastoestablisharockburstindexsystem.Then，inviewoftheproblemofdataimbalanceintherockburstdatabase，the SMOTEoversampling algorithmwasused toexpandthedatabase.Finaly，fivecommonlyused machine learning models were introducedtopredictherockburstintensitylevel，andthesefivemodelswereusedtopredictheoriginalrockburstdatabaseandthe rockburstdatabaseafterSMOTEalgorithmrespectively，toverifytheefectivenessofthepretreatmentprocess.Theresultsshow that：1）Compared with the traditional model，the prediction accuracy of the model is improved by 10.000%～35.000% after the introductionof SMOTEalgorithm；2）Comparedwiththeotherfourmodels，theSMOTE-randomforestmodelhadthehighest prediction accuracy.

Keywords：rockburst；SMOTE oversamplingalgorithm；random forest；intensitylevel prediction

来稿日期：2024-12-11 DOI：10.13937/j. cnki. hbdzdxxb.2025.03.005

基金项目：中央引导地方科技发展资金项目（246Z5405G）

作者简介：李璐佳（2000—），女，河北石家庄人，硕士研究生，主要从事地质灾害及防治方面研究。（剩余10886字）