Automatic Construction of Classifiers by Knowledge Ecosystem Agents

Abstract

Digital Ecosystem (DE) is understood as a distributed, adaptive, open socio-technical system that has the properties of self-organization, scalability and sustainability. A high level of adaptability and self-organization of DE can only be ensured by the Knowledge Ecosystem (KE) built on the principle of nesting dolls. The main purpose of KE is effective knowledge management. This is achieved as a result of improving the interaction environment for system participants, simplifying the decisionmaking process and stimulating innovations. The base elements of the Knowledge Ecosystem are software agents. They "live" in the ecosystem environment: receive and analyze data about surrounding events, interpret them and execute commands that affect the environment. The paper presents the process of automatic construction of classifiers based on the interaction of knowledge ecosystem agents. The initial information for the collaborative work of agents is the alphabet of classes, the a priori dictionary of features (PDF), and the data warehouse. The effectiveness of the proposed approach is demonstrated by the example of processing model data.