Students Classification Basing the Numerical Model to Compile Playful Teams on the Unispher™ Platform

Abstract

Methods for students classifying to compile play-learning groups are considered. Classification is necessary to create conditions for coordinated teamwork of the groups’ members. Classification is carried out according to interest and learning ability, behavioral characteristics, character traits and social characteristics. The students are presented by a vector of their features numerical assessments. Approaches are proposed based on hierarchical selection of students into groups in order to ensure equality of groups in composition and diversity of behavioral characteristics. The first approach consists in direct distribution and subsequent rotation of groups members based on characteristic histograms. The second approach uses the method of Independent Component Analysis in data driven basis obtained by Singular Value Decomposition. The method allows to classify students’ data taking into account a significance of the parameters and an uncertainty of some parameters values through their randomization and statistical analysis. In the case when groups are formed according to the criterion of the maximum possible mixing of behavioral types and characteristics of students, both methods give approximately equal results. However, the second method is more flexible and faster; it also produces clusters of students, which can be used to form groups based on other criteria.