论文部分内容阅读
将Vapnik提出的支持向量机(Support Vector Machine,SVM)算法用于总结头发中多种微量元素含量与高血压的对应关系的结果。通过对26个高血压患者和27个健康人的头发样品的多种微量元素的定量分析,用支持向量机研究头发微量元素与高血压的相关性,结果表明:若以头发中Al,Cu,Zn,Ca,Mg含量以及Zn/Cu比作为特征量集合作数据挖掘,所建数学模型对高血压患者与健康人的正确分类率可达96.2％,留一法预报正确率则可达86.7％。计算表明:支持向量机算法建模的正确分类率和留一法预报正确率均较Fisher法和KNN法等传统的模式识别算法高。因此,SVM算法是特别适合于用有限已知样本训练建模,进而预报未知样本属性的新算法,并可望在化学计量学领域得到进一步的应用。
The support vector machine (SVM) algorithm proposed by Vapnik is used to summarize the corresponding relationship between various trace elements in hair and hypertension. Based on the quantitative analysis of trace elements in hair samples of 26 hypertensive patients and 27 healthy subjects, the correlation between trace elements in hair and hypertension was studied by using support vector machine. The results showed that if hair, Al, Cu, Zn, Ca, Mg content and Zn / Cu ratio as the feature quantity set for data mining, the correct classification rate of the constructed mathematical model for hypertensive patients and healthy people can reach 96.2%, and the accuracy of leaving one method can reach 86.7% . The results show that both the correct classification rate of SVM algorithm and the accuracy of prediction of one-leave-one method are higher than the traditional pattern recognition algorithms such as Fisher method and KNN method. Therefore, the SVM algorithm is a new algorithm especially suitable for training and modeling with a limited number of known samples to predict the properties of unknown samples, and is expected to be further applied in the field of chemometrics.