A method for classifying objects with a heterogeneous set of information features based on the generation of graph-chromatic maps

Abstract

Background. In the modern world, large volumes of data are continuously processed. One type of data processing is classification. A classification task is often complicated by the need to classify objects with a heterogeneous set of features. One of the main ways to solve such problems is to reduce heterogeneous features of an object to one type. This work outlines our own original approach to solving such problems. Materials and methods. The proposed method is based on transformation objects to a universal raster representation of grapho-chromatic maps, and has a number of features that distinguish it for the better from similar methods. The possibility of using already existing convolutional neural networks as a classifier of created grapho-chromatic maps is considered. Results and conclusions. A block diagram of a system for classifying objects using the proposed method has been developed. Each element of the system is described in detail. An algorithm for generating a grapho-chromatic map is proposed. Conclusions are drawn about the potential capabilities and advantages of the proposed approach.

Keywords

• classification
• methods of classification
• heterogeneous data
• machine learning
• neural networks
• graph-chromatic map