CASPIAN JOURNAL

MANAGEMENT AND HIGH TECHNOLOGIES

NEURAL NETWORKS COLLECTIVE MODEL AND SOFTWARE PACKAGE TO DETERMINE PERSON'S FUNCTIONAL STATE

Read

Akhmetvaleev Amir M, Katasev Alexey S., Podolskaya Marina A. NEURAL NETWORKS COLLECTIVE MODEL AND SOFTWARE PACKAGE TO DETERMINE PERSON'S FUNCTIONAL STATE // Caspian journal : management and high technologies. — 2018. — №1. — pp. 69-85.

Akhmetvaleev Amir M - postgraduate student, Kazan National Research Technical University named after A.N. Tupolev - KAI, 10 K. Marx St., Kazan, 420111, Russian Federation, amir.akhmetvaleev@gmail.com

Katasev Alexey S. - Cand.Sci. (Engineering), Associate Professor, Kazan National Research Technical University named after A.N. Tupolev - KAI, 10 K. Marx St., Kazan, 420111, Russian Federation, kat_726@mail.ru

Podolskaya Marina A. - Cand.Sci. (Medicine), Kazan State Medical Academy, 36, Butlerova St., Kazan, 420012, Russian Federation, maro7@mail.ru

The paper presents the urgency of solving the problem of determining person's functional state (on the example of diagnostics of the state of intoxication). Relevance of using an approach based on application of the method of pupillometry is proved. It allows to evaluate person's state by his pupillary response to changing illumination. There is the statement of the problem of determining whether person is in the state of intoxication by the results of pupillogramm parameter values analysis for time series, characterizing changing in the size of a pupil when it is exposed by a short-term light impulse. It is proposed to use mathematical tool of artificial neural networks to analyze pupillogramms. Characteristics of the source data for neural network model construction are given. Stages of collecting and preparing data for analysis; developing a primary neural network structure; determining optimal point of cutting-off sets of decisions for the model; using bootstrap service to construct the original neural network model as a set of neural networks; structural model optimization on the base of developing and using two-stage genetic algorithm are described. Instrumental set of programs is developed to automate process of determining person's functional state. Its structure, composition, component characteristics, as well as the results of the research are presented. On the basis of the developed mathematic and software a general scheme of determining person's functional state is proposed. The efficiency of using methods based on this scheme, when performing pre-trip medical examination of drivers, is proved.

Key words: функциональное состояние, пупиллометрия, модель коллектива нейронных сетей, ROC-анализ, бутстрэпирование, редукция, генетический алгоритм, предрейсовый медицинский осмотр, functional state, pupillometry, collective neural network model, ROC analysis, boot