Wavelet neural network and its application to the inclusion of β-cyclodextrin with benzene derivatives

A wavelet neural network (WNN) was constituted and applied to the inclusion complexation of beta-cyclodextrin with mono- and 1,4-disubstituted benzenes. The association constant (K-a) values have been calculated by the WNN from substituent molar refraction (R-m), hydrophobic constant (pi), and Hammett constant (sigma) of the guest compounds as input parameters. The excellent prediction results with a correlation coefficient of 0.992 and standard deviation of 0.089 suggested that beta-CD inclusion complexation is mainly driven by van der Waals force, hydrophobic interaction, and electronic effects.