BP人工神经网络法研究PBDE_S类化合物的RRT

摘要：

运用BP人工神经网络方法对PBDEs的相对保留时间(RRT)进行了QSPR研究.所建的BP人工神经网对PBDEs的RRT预测准确度非常高,网络训练误差几乎为0,网络回判MSE误差为0.003 9,明显低于逐步回归分析结果,独立检测集MSE误差为0.000 4,也很低,说明BP人工神经网具有较好的泛化能力.此方法得到的模型预测能力要优于逐步回归模型.

QSPR studies on RRT of PBDEs were carried out by BP artificial neural network method. The prediction accuracy of the BP artificial neural network on PBDEs RRT was very high. The network training error was almost 0. The network predates MSE error was 0. 003 9 which was significantly lower than the results of stepwise regression analysis. Detection of MSE in the independent predictive error was 0. 000 4 , and it was very low. The results showed the generalization ability by BP artificial neural network was better, and the model constructed by this method had better prediction ability than the model of stepwise regression model.

作者：

曹红翠 孙海霞 保英莲

机构地区：

青海大学化工学院

出处：

《betway官方app 学报：自然科学版》 CAS 北大核心 2015年第1期84-87,共4页

基金：

教育部春晖计划(Z2012085)

关键词：

多溴联苯醚(PBDEs) BP人工神经网络 RRT

Polybrominated diphenyl ethers （PBDEs） BP artificial neural network RRT

分类号：

O641.12 [理学—物理化学]

BP人工神经网络法研究PBDE_S类化合物的RRT.pdf