基于3种机器学习模型的岩爆类型预测

Rockburst type prediction based on three machine learning models

基于国内外397组岩爆案例数据,采用模型参数优化及交叉验证技术获得最近邻、支持向量机与决策树模型最佳参数;对比主成分分析(PCA)与过采样SMOTE(synthetic minority oversampling technique)算法对3种机器学习算法预测准确率的影响,并对模型预测性能进行评估.结果表明:主成分分析对3种机器学习模型的预测准确率并无提升,不同岩爆类型的样本之间不具有较为明显的分类边界;过采样SMOTE算法仅对决策树模型有明显的提升,基于过采样建立的SMOTE-DT模型预测准确率为77.50%,高于仅对原始数据集进行标准化处理的K-最邻近(KNN)和支持向量机(SVM)模型的68.75%与57.50%;SMOTE-DT在避免高估与低估岩爆类型表现优于KNN与SVM模型,对于4种岩爆类型的F_1值均大于0.7,岩爆预测性能稳定可靠.此外,采用本文构建的3种机器学习模型对山西紫金金矿进行岩爆类型预测,模型预测结果与现场观测结果相一致.

Based on 397 groups of rockburst case data,the best parameters of the nearest neighbor,support vector machine and decision tree model are obtained by using model parameter optimization and cross validation techniques.Compare the influence of PCA(principal component analysis)and oversampling SMOTE(synthetic minority oversampling technique)on the prediction accuracy of three machine learning algorithms,and evaluate the prediction performance of the model.The results showed that principal component analysis did not improve the prediction accuracy of the three machine learning models,and there was no clear classification boundary between samples of different rock burst types.The oversampling SMOTE algorithm only significantly improves the decision tree model.The prediction accuracy of the SMOTE⁃DT model based on oversampling is 77.50%,which is higher than 68.75%and 57.50%of the K-nearest neighbor(KNN)and support vector machine(SVM)models that only standardize the original data set.SMOTE⁃DT performs better than KNN and SVM models in avoiding overestimation and underestimation of rock burst types.The F1 values for all four rock burst types are greater than 0.7,indicating stable and reliable rock burst prediction performance.In addition,the three machine learning models constructed in this article were used to predict the type of rockburst in Shanxi Zijin Gold Mine,and the predicted results of the models were consistent with the on⁃site obser⁃vation results.