摘要
将自适应映射 (SOM)用于多环芳烃致癌性的分级。采用的输入参数为分子比表面积、代谢活性区及亲电活性区的中心碳原子离域能、分子中脱毒区总数。优化的网络参数包括网格数及网格形状、学习次数和学习率、邻居半径等。在最佳网络参数下 ,多环芳烃致癌性分类准确度大于 97%。
The self organizing map (SOM) was used in the visual classification of the carcinogenicity of 77 polycyclic aromatic hydrocarbons (PAHs). They were classified based on the following attributes: the total surface area (TSA) of the molecule, the delocalization energy of the center carbon atom in the metabolic active region (Δ E 1), the delocalization energy of the center carbon atom in the electrophilic active region (Δ E 2), and the number of de toxic regions in the molecule ( N d). The classification of SOM in this field is very satisfactory.
出处
《分析化学》
SCIE
EI
CAS
CSCD
北大核心
2000年第11期1336-1343,共8页
Chinese Journal of Analytical Chemistry
基金
甘肃省自然科学基金!(ZS981 A2 5 0 51C)资助课题
关键词
神经网络
自适应映射
多环芳烃
致癌性
分类
Neural network, self organizing maps, polycyclic aromatic hydrocarbons, carcinogenicity
