Treffer: Multiagent System for Monitoring, Analysis and Classification of Data from Procurement Services.

The article describes the method for automatization of procurement notices (tenders) monitoring, analysis and classification. This method is aimed at simplification and reduction of labor costs for search, gathering and subsequent analysis of tenders, as public procurement has become one of the major sources of new projects. The method describes a multi-agent system for automated extraction, processing, classification, storage and visualization of procurement notices data. Agents use open-source and SSPL-based solutions. Having keywords, publication date and other requirements as input, the program returns collection and visualization of tenders from five marketplaces as output. Output is given in Excel table, index in Elasticsearch database and visualization in Kibana (maps and plots) and Neo4j (graphs), thus providing a user with an instrument for automated extraction and analysis of tender data. Classification is carried out via machine learning technologies provided by Python scikit-learn module. Classification model has shown an efficiency of 80-85%. The article describes the method itself, the method performance on tenders related to decommissioning in nuclear sphere, and prospects of method. [ABSTRACT FROM AUTHOR]