用BP神经网络法预测反相液相色谱保留值

Prediction of retention indices in reversed-phase high-performance liquid chromatography with BP artificial neural networks

运用逐步回归法筛选BP人工神经网络的输入元素改进BP网络的结构 ,并用于疏水分配常数及氢键作用能预测反相液相色谱保留值 .结果表明 ,逐步回归法与人工神经网络相结合 ,能优化网络 ,大大提高网络收敛速度 ,在很大程度上克服了通常BP网络过拟合的缺点 .较通常BP神经网络及回归法在处理构效关系方面有更强的信息处理能力和预测能力 .

The back-propagation (BP) neural network is improved by combining the multivariable stepwise regression and BP neural network,and used to predict the rentention indices in reversed-phase high-performance liquid chromatography with the partition coefficient and hydrogen bonding ability.The results show that the improved BP neural network has better performance such as accelerating the convergence and improving predictable ability than the original one and multivariable regression method.