Matvieiev O. Multi-agent models and methods of collecting and extracting business information in the web space

The scientific and practical task of collecting and extracting business information in a distributed information environment to ensure the relevance and effectiveness of the decisions made is solved in the dissertation work. The object of research is the process of collecting and extracting information in the web space. The subject of research is models and information technology for collecting and extracting relevant data in the distributed information space in order to form informed decisions. The following methods were used to solve the problems: system analysis for the development of a conceptual model of web monitoring; methods of the theory of artificial intelligence and the algebra of finite predicates of the first order to create a formal architecture of agents; the method of comparative identification, methods of intelligent data analysis, methods of building ontologies and methods of mathematical statistics for processing data collected from web pages; methods of decision-making theory; the unified UML modeling language and agent-oriented programming methodology for the implementation of a prototype of an agent platform based on the use of formal methods for designing multi-threaded systems. In the first section of the dissertation, an analysis of information search and monitoring systems, approaches to their solution, and an analysis of the shortcomings of the existing thematic search apparatus were carried out. The problems that have not been solved at the moment and are relevant from the point of view of analysis and increasing the efficiency of business decision-making are highlighted. The direction of increasing the efficiency of decision-making support processes due to the collection of relevant data in the web space has been determined. The existing methods and means of an integrated approach to the implementation of the web space data search process based on intelligent technologies are considered. The most popular and widely used programming paradigms are analyzed. In this work, it is proposed to use an agent architecture when designing a monitoring system. In the second chapter of the dissertation, the theoretical foundations of improving the efficiency of business decision-making by developing multi-agent models and methods of collecting and extracting business information in the web space are formed. The use of the mathematical apparatus of algebra of finite predicates is substantiated, the models of search, extraction and extraction of information based on the use of a comparative model are developed. The concept of collecting and extracting business information based on the use of multi-agent systems is considered. The paper proposes the use of FIPA standards, which provide basic definitions of the concepts of agent communication. The following new scientific results were obtained on the basis of the research carried out, reflected in this section: a comprehensive approach to the implementation of the process of data collection of the web space based on intelligent technologies was formed; the business information extraction model has been improved. In the third chapter, the methods of searching and collecting business information are improved by using the method of comparative identification to build a model for evaluating the relevance of a web page and a multi-agent model of directed search, which provides the opportunity to search for information resources based on the principle of usefulness for decision-making. The research of this section made it possible to obtain the following results: the model of interpretation of incomplete contradictory information obtained by collecting and extracting it in the web space was further developed. The fourth chapter of the dissertation presents the main results regarding the development of information technology for the collection and extraction of business information based on the proposed method and the developed models for the extraction, interpretation and evaluation of data contained in the web space through the implementation of a multi-agent data processing model. In this work, a prototype of an agent platform for the implementation of multi-threaded systems in various subject areas has been developed. The paper presents the results of experiments that prove the efficiency of the proposed models. According to the studies of this section, the following results were obtained: a specification of software requirements was created; software components of the server part of the software system were developed, which allows data extraction from virtual trading platforms; conducted experiments on the suitability of the agent platform for solving the task of collecting and extracting business information; testing of the developed software components was carried out and their effectiveness for the implementation of multi-agent data processing was proven.