人工神经网络预测阴离子表面活性剂的临界胶束浓度

Prediction of critical micelle concentration of anionic surfactant by artificial neural network

根据定量结构-性能关系原理,利用人工神经网络模型,采用Wn、S和Sn三种拓扑指数描述阴离子表面活性剂的分子结构并作为网络输入,预测阴离子表面活性剂的临界胶束浓度。确定了网络参数,利用43组数据对网络进行训练和预测,并与文献值进行了比较。结果表明,预测精度较高,说明人工神经网络方法具有很好的预测能力。

Based on quantitative structure- property relationship （QSPR）, utilizing artificial network, three topological parameters W,, S and S, were used to describe molecular structure of anionic surfactants and as inputs to the network to predict critical micelle concentration of anionic surfacvant. The parameters of network were identified. Altogether, there were 43 groups of data applied for training and prediction. Results were compared with that proclaimed in literatures and showed that the prediction was very accurate and the network approach is effective in this application.