火山岩的形成环境复杂,有些地区的火山岩可能只发育两三种岩石类型,这会导致不同岩性取心资料的代表性严重失衡.针对现有的测井岩性识别方法在处理类间不均衡样本时出现效果较差的问题,提出基于ADASYN-GS-XGBOOST混合模型的火山岩岩性识别方法.首先通过ADASYN过采样算法对不均衡样本进行处理得到新的样本集,再以XGBOOST算法作为基分类器对样本进行分类,并利用网格搜索法(GS)对模型进行参数优化,以此建立ADASYN-GS-XGBOOST混合岩性识别模型.将该混合模型训练后的结果与K近邻、朴素贝叶斯、随机森林、XGBOOST及SMOTE-GS-XGBOOST等算法的岩性识别结果进行对比,表明基于ADASYN-GS-XGBOOST算法建立的模型识别效果最好.该方法克服了已有岩性识别方法无法有效解决不均衡样本的问题,极大地提升了火山岩岩性识别的准确率.

The forming environment of volcanic rocks is complex,and the lithology of volcanic rocks in a certain area may be mainly composed of two or three types,which leads to serious imbalance of core data of different lithology.The existing lithology identification methods are not effective in dealing with unbalanced samples among classes.To solve these problems,a volcanic rock lithology identification method based on ADASYN-GS-XGBOOST hybrid model is proposed.The unbalanced samples are processed by ADASYN oversampling algorithm to obtain a new sample set,and then XGBOOST is used as the base classifier to classify the samples.The ADASYN-GS-XGBOOST hybrid lithology identification model is established by using Grid Search to optimize the parameters of the model.The results of the hybrid model training are compared with those of K nearest neighbor,naive Bayes,random forest,XGBOOST and SMOTE-GS-XGBOOST algorithms.The results show that the model based on ADASYN-GS-XGBOOST algorithm has the best identification effect.This method overcomes the problem that existing lithology identification methods can not effectively solve the problem of unbalanced samples,and greatly improves the accuracy of lithology identification of volcanic rocks.