Content adaptation neural network method cause-specific the state of users

Abstract

The aim of the study is to develop recommendations for adapting educational content based on the use of a neural network method for recognizing the human autonomic nervous system degree of activity. The initial data for decision making are the vectors of cardiointervals obtained with the help of the pulse sensor. The state of the autonomic nervous system is monitored by a two-layer artificial neural network of direct propagation. The artificial neural network was trained by combining gradient and stochastic training methods. A trial sample was used as a training set, consisting of 168 records of people’s cardiointervals belonging to different social and age groups. The presented method for recognizing the autonomic nervous system activity degree was tested via the shared programming support environment Lazarus. As a result of network training, a set of weighting coefficient values was found, for which 100% correct recognition is performed over the entire trial sample. A personalized training system that monitors the functional state of the student varies the presentation of educational content. When the functional state changes during the training load, the virtual agent organizes the training process so as to avoid stress load.