Alexander Nikolaevich Dmitriev - Diploma student, Chair «Medical and technical information technologies», Faculty «Biomedical Engineering», Bauman Moscow State Technical University. E-mail: dmitalexnic@gmail.com
Vladimir Vladimirovich Kotin - Ph.D. (Phys.-Math.), Assistant Professor, Chair «Medical and technical information technologies», Faculty «Biomedical Engineering», Bauman Moscow State Technical University. Е-mail: v.kotin@gmail.com

One of the essential problems of nowadays is developing new practically demanded models and methods of social science and mathematical epidemiology and their integration into the sphere of solving medical problems. The aim of such research is raising the efficiency of planning the events to response to epidemics and improving the population dynamics. Methods of computational and artificial intelligence can be applied to solve the problems of disease prediction. The use of fuzzy sets for the prediction of disease dynamics allows to improve the validity and controllability of the management decision-making with the threat of epidemics. Hybridization, or in other words, development and application of ensembles of models that interact with the aim of mutual compensation of disadvantages and limitations of each individual model, is an approach to solving a wide range of forecast problems of the incidence. Fuzzy Systems with the paradigm of hybridization are used to analyze the data together with other methods of intelligent computing, such as multi-agent simulation, artificial neural networks, genetic algorithms, expert systems, etc., forming a useful alternative to dynamic models. The main types of fuzzy models are Mamdani and TSK (Takagi-Sugeno-Kang), they have a modular structure, ideal for system representation in the form of a uniform multilayer structure resembling the structure of the classical neural networks. The package MATLAB Fuzzy Logic Toolbox neuro-fuzzy network are implemented in the form of adaptive neuro-fuzzy inference ANFIS (Adaptive Neuro-Fuzzy Inference System). The model ANFIS is taken as the basic structure of the neuro-fuzzy system. To analyze the possibilities of fuzzy neural network prediction of disease, the data of the incidence of scarlet fever in Moscow in 1996-2008 was taken. Tests showed that the average relative error of prediction is 37%.