Improving microseismic event and quarry blast classification using Artificial Neural Networks based on Principal Component Analysis
Improving microseismic event and quarry blast classification using Artificial Neural Networks based on Principal Component Analysis
复制标题
使用基于主成分分析的人工神经网络改进微震事件和采石场爆炸分类
DOI：
10.1016/j.soildyn.2017.05.008
复制
发表时间：
2017-08-01
影响因子：
4
通讯作者：
Chen, Guanghui
中科院分区：
文献类型：
--
作者：
Shang, Xueyi;Li, Xibing;Chen, Guanghui
本文研究了微震事件与采石场爆破的区别。为此，使用了主成分分析(PCA)和人工神经网络(ANN)。对1600个地震事件的22个地震参数进行了检验。在这项工作中，主成分分析被用来将原始数据集转换为一个新的不相关变量数据集。生成的新数据集已用作ANN的输入，并与Logistic回归(LR)、贝叶斯和Fisher分类器进行比较，后者对微震事件和采石场爆炸进行分类。结果表明，主成分分析对评价变量和数据降维是有效的。此外，基于主成分分析的分类结果要好于基于文献[1]的分类结果。[22]并且不使用主成分分析方法。而且，神经网络分类器的分类效果最好。主成分分析的马修相关系数(MCC)结果，参考文献[22]而不使用主成分分析的方法分别达到了89.00%、73.68%和82.04%，显示了基于主成分分析方法的可靠性和潜力。
The discrimination of microseismic events and quarry blasts has been examined in this paper. To do so, Principal Component Analysis (PCA) and Artificial Neural Networks (ANN) have been used. The procedure proposed has been tested on 22 seismic parameters of 1600 events. In this work, the PCA has been used to transform the original dataset into a new dataset of uncorrelated variables. The new dataset generated has been used as input for ANN and compared to Logistic Regression (LR), Bayes and Fisher classifiers, which classify microseismic events and quarry blasts. The results have shown that PCA is effective for rating variables and reducing data dimension. Furthermore, the classification result based on PCA has been better than those based Ref. [22] and without PCA methods. Moreover, the ANN classifier has obtained the best classification result. The Matthew's Correlation Coefficient (MCC) results of the PCA, Ref. [22] and without PCA based methods have reached 89.00%, 73.68% and 82.04%, respectively, thus showing the reliability and potential of the PCA based method.