饱和醇定量结构-保留相关研究中人工神经网络的应用

The Application of Artificial Neural Networks in the Study of Quantitative Structure-Retention Relationships for Saturated Alcohols

以拓扑指数（分子连接性指数）为结构描述符，用人工神经网络技术建立了醇类化合物的结构与色谱保留值的相关性模型。研究了网络构造对模型稳定性的影响，考察了模型在单一固定相上及多固定相上的适应性。与多元线性回归法相比较，人工神经网络模型具有更好的预测结果，但外推能力较弱。

A model of artificial neural network (ANN) about the relationship between the structure of alcohols and their chromatographic retention indices has been set by means of topological indices. The influence of the network structure for the stability of the model has been studied, and the compatibility of the model for mono-stationary phase and multiple stationary phases has also been investigated. Compared with multiple linear regression (MLR) method, the model of ANN is better than MLR in prediction, but worse in extrapolated ability.