NEURAL-NETWORK STUDIES .2. VARIABLE SELECTION

Quantitative structure-activity relationship (QSAR) studies usually re quire an estimation of the relevance of a very large set of initial va riables. Determination of the most important variables allows theoreti cally a better generalization by all pattern recognition methods. This study introduces and investigates five pruning algorithms designed to estimate the importance of input variables in feed-forward artificial neural network trained by back propagation algorithm (ANN) applicatio ns and to prune nonrelevant ones in a statistically reliable way. The analyzed algorithms performed similar variable estimations for simulat ed data sets, but differences were detected for real QSAR examples, Im provement of ANN prediction ability was shown after the pruning of red undant input variables. The statistical coefficients computed by ANNs for QSAR examples were better than those of multiple linear regression . Restrictions of the proposed algorithms and the potential use of ANN s are discussed.