中文摘要：

SVC和SVR是支持向量机研究的两个主要问题。文章把两种建模方法相结合,先由SVC模型判别分类,后由各类的局部SVR模型进行定量分析,提出了复合支持向量机（CSVM）方法。根据71个试验小区的水稻冠层高光谱与叶片含氮量建立定量分析模型,考证了CSVM算法。基于模拟研究的思想,随机划分建模集和预测集,比例为55∶16。经过5次划分试验,复合支持向量机方法建模对叶片含氮量的预测值与凯氏定氮实际值之间的平均相关系数为0.89,平均绝对误差为0.088;而传统的支持向量机方法得到的平均相关系数为0.87,平均绝对误差为0.091。由此可见,复合支持向量机方法相对于传统的支持向量机方法预测精度有所提高。文章研究方法的提出为化学计量学定量分析研究给出了新的思路。

英文摘要：

Support vector classification （SVC） and support vector regression （SVR） are two main issues of support vector machines （SVM）. The present paper combined the two issues, that is, first to built SVC model for classification, then to built SVR models for analysis, and thus brought forward compound support vector machines （CSVM）. Based on an idea of simulation study, the CSVM algorithm was built and then validated by building a quantitative analysis model using high-spectrum and leaf nitrogen content data of 71 rice samples which were divided into modeling set and forecasting set randomly at the ratio of 51 ： 16. For 5 random experiments, the average correlation coefficient of predicted values and standard chemical ones by Kjeldahl＇s method of leaf nitrogen content was 0. 89, and the average absolute error was 0. 088, of which the corresponding values produced by traditional method were 0. 87 and 0. 091 respectively. It was concluded that the prediction precision of CSVM is higher than that of traditional SVM. CSVM provides a new idea for chemometrics quantitative analysis.