核主成分分析-广义回归神经网络法研究——70种烷烃类有机化合物的气相色谱保留指数  

Study on Gas-chromatographic Retention Index of 70 Alkanes by Kernel Principal Component Analysis-generalized Gegression Neural Network Method

作　　者：许国根[1] 徐昊[1] 陈然[1] 

机构地区：[1]第二炮兵工程学院,陕西西安710025

出　　处：《化工技术与开发》2012年第1期26-31,共6页Technology & Development of Chemical Industry

摘　　要：利用核主成分分析法和广义回归神经网络研究了70种烷烃在固定相为角鲨烷,柱温50℃时的气相色谱保留指数。核主成分分析将70×8维的样品集降为51×3维,以减少计算量;降维后的样品集作为广义回归神经网络的训练集对色谱保留时间进行计算和预测。该方法的计算结果满足误差要求。A method for the estimation and forecasting of the gas-chromatographic retention indexes of 70 alkane on squalane stationary phase at the column temperature of 50℃ was developed by means of the kernel principal component analysis-generalized gegression neural network.The kernel principal component analysis is used to reduce 70×8 dimensional samples to 51×3 dimensional samples in order to save the calculation time and then the generalized regression neural network is used to predict the retention indexes of 19 alkanes using 51×3 dimension samples as training cluster.The calculation results of the method meet the error demand.

关 键 词：核主成分分析 广义回归神经网络 色谱保留指数 

分 类 号：O657.71[理学—分析化学]