论文部分内容阅读
子结构模式识别一直是低分辨率质谱中的难点问题,为提供更多的分类信息,达到更好的识别和分类效果,将质谱图中被脱去碎片的信息,通过峰簇运算的方法提取出来,建立了一种新的质谱分类变量:FM(Fractional Mass)。联合块变量正交化和典型相关分析方法,以饱和醇醚的模式识别模型对变量进行了验证。结果表明,分类效果较好,无论是总体的模型还是交互检验,预测误差率都在2.5%以下,FM变量是传统分类变量的有益补充。
Substructure pattern recognition has always been a difficult problem in low-resolution mass spectrometry. In order to provide more classification information and achieve better recognition and classification effects, the extracted fragment information in the mass spectrum is extracted by using a cluster computing method Come out, established a new classification of mass spectral variables: FM (Fractional Mass). Combined with the orthogonalization of block variables and canonical correlation analysis, the variables were validated by the pattern recognition model of saturated alcohol ether. The results show that the classification effect is good, both the overall model and the interactive test, the prediction error rate is below 2.5%, FM variable is a useful complement to the traditional classification variables.