【摘要】 采用Chemoffice10.0中的MOPAC/AM1半经验算法,计算13个无机氢化物的6种量子化学参数:偶极距(μ)、分子电子能(EE)、分子最高占据轨道能(E homo)、分子最低未占据轨道能(E lumo)、分子总能量(TE)和非氢原子的净电荷(q),利用SPSS19.0软件的主成分分析模块和BP神经网络模块对样本数据集进行预处理,将得到的样本数据集输入人工神经网络,构建无机氢化物pK a的主成分-反向传播神经网络(PCABPNN)QSPR模型.结果表明:利用SPSS19.0软件快速实现PCA-BPNN模型取得了非常满意的结果,而且模型的相关系数高,所得结果均优于传统的多元线性回归方法.更多还原

【Abstract】 The thesis calculates 6 kinds of quantum chemistry parameters of 13 inorganic hydride compounds by using the MOPAC / AM1 method in Chemoffice10. 0: dipole moment(μ),electronic energy( EE),energy of the highest occupied moleular orbit( E homo),energy of the lowest unoccupied moleular orbit( E lumo),total energy of molecule( TE) and the net atomic charges on atom of nonhydrogen( q). The sample data set is preprocessed by principal component analysis module and neural network module in SPSS19. 0. The new principal components analysis( PCA) was used as input parameters of BPNN analysis to predict pK a of 13 inorganic hydride compounds,and the QSPR model was established. The results show that using SPSS19. 0 software to realize PCA-BPNN prediction has achieved satisfactory and the model has high significant correlation coefficient. The results of BPNN are better than the traditional multiple linear regression methods.更多还原