基于支持向量机和有序聚类的岩层识别

Stratum identification based on the SVM and ordered cluster

由于支持向量机进行分类前需要先使用训练样本训练分类器,而在岩层识别问题中没有训练样本,针对此问题,提出一种基于有序聚类的支持向量机岩层识别分类算法。首先利用有序聚类算法对经滤波和归一化后的测井数据进行初步分层,然后根据初步分层结果获取训练样本,最后用训练后的支持向量机分类器对测井数据进行第2次分层。应用该算法对选取的3口井的岩性进行自动识别,并将该算法的识别结果与其他算法进行比较。仿真实验结果表明,该算法具有较高的准确率,每种岩层的平均准确率能达到85%,解决了岩层识别前必须采用已知类别的数据对支持向量机进行训练的弊端。

The support vector machine （ SVM ） needs training samples to train itself before identifying stratum , while there are no training samples with stratum identification .Focusing on this problem , this paper puts forward a vector machine classifier based on the ordered clustering algorithm .Firstly, the ordered clustering algorithm is used to get preliminary layered logging data which have been filtered and normalized .Secondly , the training samples are obtained according to preliminary layered outcomes .Finally, the data are layered again by the trained SVM classifi-er.The algorithm is used to automatically identify the lithology of the selected three wells , and compared with the results of the other algorithms .The results of the simulation experiment show that the algorithm overcomes the draw-backs that the labeled data has to adopt when training SVM , and improves the accuracy of each stratum , reaching 85%on average .