Comparison of conventional artificial neural network and wavelet neural network in modeling the half-wave potential of aldehydes and ketones

A Quantitative Structure-Electrochemistry Relationship (QSER) study has been done on the half-wave reduction potential (E_1/2) of some organic compounds containing 73 aldehydes and ketones using multiple liner regression (MLR), partial least square (PLS), artificial neural network (ANN) and wavelet neural network (WNN) modeling methods. First, stepwise multiple liner regression was employed as a descriptor selection procedure. Then selected descriptors were used as inputs for artificial neural network and wavelet neural network models. In this paper we have studied the abilities of conventional ANN and WNN for prediction of half-wave potential (E_1/2) of aldehydes and ketones. Comparison of the results indicates that the ANN and WNN as nonlinear methods have better predictive power than the linear methods. The stability and prediction ability of these models were validated using 10-fold cross-validation, external test set, and Y-randomization techniques.