1. Использование нейронных сетей в задаче выявления доминантных личностных характеристик различных категорий респондентов
- Subjects
ПСИХОДИАГНОСТИКА,ЛИЧНОСТНАЯ ХАРАКТЕРИСТИКА,КЛАССИФИКАЦИЯ,КЛАСТЕРИЗАЦИЯ,ПЕРСЕПТРОН,СЕТЬ КОХОНЕНА,СЛОЙ ГРОССБЕРГА,СЕТЬ ВСТРЕЧНОГО РАСПРОСТРАНЕНИЯ,PSYCHODIAGNOSTICS,PERSONAL CHARACTERISTIC,CLASSIFICATION,CLUSTERIZATION,PERCEPTRON,KOHONEN NETWORK,GROSSBERG LAYER,COUNTER PROPAGATION NETWORK - Abstract
Рассмотрена возможность применения аппарата нейронных сетей для решения задачи классификации в области психодиагностики. В качестве исходных данных используются результаты тестирования заключенных по методике Лири. В статье приводится структурно-функциональная схема созданного программного комплекса, описываются этапы его функционирования, виды нейронных сетей, используемых на каждом из этапов работы комплекса., In the article the description of research papers on the applicability check of neural network technology to the problem of the personal characteristics is presented. The personal characteristics define the distinction in psychologic portraits of various categories of respondents (dominant personal characteristics). As initial data for carrying out the research, the results of psychodiagnostic testing of respondents by means of Leary's procedure were taken and the toolkit is comprised of the perceptron neural network and the network of counter propagation combining the Ko-honen network and the Grossberg layer. Research includes two stages. Stage 1. Classification. The problem of detection possibility of a set of personal traits which could act as a basis for classification is solved. The research was carried out with the use of a perceptron neural network. As combinations of personal characteristics which could be a basis of possible classification are unknown in advance, it is necessary to realise a perceptron cycle on sets of samples from primary tests, and the measurements of input vectors, as a rule, influence on the perceptron structure. Therefore, for automation of process at the given stage (preparation of input data sets, neural network generation and training, and classification results display) the specialised program complex is created. Stage 2. Clusterization. At the second stage (provided that the positive results of classification are obtained) by means of a counter propagation network the clusterization problem is solved, which results could confirm or deny the results of classification. The examples of research results allow considering the described approach to the solution of a certain class of problems of psycho-diagnostics as perspective enough and capable to interest practitioners of the corresponding directions of psychodiagnostics.
- Published
- 2016