ОЦЕНКА ЭФФЕКТИВНОСТИ ПРИМЕНЕНИЯ ИСКУССТВЕННЫХ НЕЙРОННЫХ СЕТЕЙ В МЕДИКО-ЭКОЛОГИЧЕСКИХ ИССЛЕДОВАНИЯХ

In a scientifi c article contains material studies on the effectiveness of the use of artifi cial neural networks in medical and environmental research. We studied the following types of neural network models: based on multilayer perceptron (MLP), radial basis function (RBF) and generalized regression network (GRNS). The control group used the linear model. Were received and analyzed 92 neural network models, 20 of them GRNS, 30 RBF, 31 MLP and 11 linear models. Evaluating the effectiveness of neural network models based on the following parameters: performance model, the magnitude of the error on the test sample, the ratio of the standard deviation (SD) of the prediction error and the original data, as well as the Pearson correlation between the observed and predicted by the model parameters. Found that linear models have a low level of effi ciency in predicting the spread of disease. Among the studied neural network models have the highest quality prediction model based on generalized regression neural networks, and especially – based networks using radial basis functions. Quality indicators in predicting neural network models of each species (GRNS, RBF and MLP) are variable enough that requires careful selection of the most effective networks.