Recognition models based on support elementary classifiers
Abstract
The issue of identifying objects defined in a high-dimensional feature space is considered which is also widely used in economics. An original approach to constructing a recognition model (RM) based on the formation of support elementary classifiers is proposed. The main idea of the proposed approach is to construct elementary classifiers based on the assessment: 1) interrelationships between features; 2) proximity between objects. A characteristic feature of the proposed model is the selection of a set of characteristic features and the formation of preferred support elementary classifiers when constructing RM on support elementary classifiers. The construction of elementary classifiers is carried out in the form of threshold functions (TF). The main advantage of the proposed RM is the increase in accuracy and reduction in the volume of computational operations when recognizing objects of the control sample, which allows it to be used in the construction of recognition systems operating in real time.