Fuzzy classification and fast rejection rules in the structure-property problem

A new approach to analysis of the molecule-descriptor matrix in the structure-property problem, based on the fuzzy cluster structure of the training sample, is developed. Methods for constructing fast prediction rejection rules and for the search of outliers in a training sample are described. To that end, a special space of easily computed descriptors is introduced. Optimization of the classifying function with respect to the parameters of fuzzy classification is considered. Prognostic models with a high quality of prediction, based on this approach, are proposed. Comparison of models is performed, which shows the efficiency of the described methods.