论文部分内容阅读
为解决在有限监测信息下,矿井突水水源识别中原始指标众多,识别矩阵呈现出的高维、稀疏性问题,采用相关性理论和支持向量机(SVM),提出了矿井突水水源识别方法.计算训练样本相关性及单指标误判个数.依据误判数对各指标进行重要性排序.构建指标组合并约简出测试正确率最高的指标组合,再利用构建的基于相关性理论的矿井突水水源SVM识别模型,实现高维小样本评价.结果表明:Ca~(2+),Mg~(2+),K~++Na~+彼此间相关性较高;K~++Na~+和CL~-,HCO_3~-和SO_4~(2-)相关性均较低;Ca~(2+)较重要,SO_4~(2-)和CL-较不重要;约简评价体系的冗余指标,可节约监测成本,且能够在有限监测信息下,保证识别正确率.
In order to solve the high initial dimension and sparsity of identification matrix in mine water inrush water source identification under the limited monitoring information, a correlation method and support vector machine (SVM) are proposed to identify water inrush source .Correlation of training samples and the number of single misjudgment.According to the misjudgment, the importance of each index is ranked.Considering the combination of indicators and reducing the combination of the indicators with the highest test accuracy, and using the constructed correlation theory The results show that there is a high correlation between Ca ~ (2 +), Mg ~ (2 +) and K ~ ++ Na ~ +; K ~ ++ The correlations between Na ~ + and CL ~ -, HCO_3 ~ - and SO_4 ~ (2-) are low, Ca 2+ is more important, SO 4 2- and CL- are less important, and the reduction evaluation system Of redundant indicators, can save the cost of monitoring, and can be limited monitoring information to ensure that the recognition rate.