Mineração de dados abertos: uma análise do uso de bots em pregões eletrônicos

• Autor(a) principal: Souto, Hugo Medeiros
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFPB
• Texto Completo: https://repositorio.ufpb.br/jspui/handle/123456789/19527
• topic: Governo aberto; Mineração de dados; Pregão eletrônico; Bots; Gestão da segurança da informação; Open government; Data mining; Electronic reverse auction; Information security management; CNPQ::CIENCIAS HUMANAS::EDUCACAO

Brazil's federal government has sought to match procurement procedures to trends in information and communication technologies. The electronic reverse auction was one of the products of these efforts, being characterized as a modality that presented structural solutions to improve the efficiency of purchases of common goods and services and that represents more than 94% of the bids that occurred in the country. Despite the benefits of electronic format, this environment brings challenges, such as dealing with the use of bots, which works by automatically bidding. While there is no law prohibiting its use, judgments of the Federal Court of Auditors state that its use provides a competitive advantage to suppliers holding this technology in question over other bidders, characterizing an affront to the principle of isonomy. Also in the sense of modernizing public procurement is increasing transparency through open data policies, as part of the context of Open Government and digital transformation. This study aims to analyze the situation of bot use in electronic reverse auctions through open data mining. Electronic reverse auctions held at the Ministry of Agriculture, Livestock and Supply in 2017 were analyzed. Data were obtained by request by the Electronic Information System for Citizen Information (e-SIC), having been adopted as methodology the knowledge discovery in databases. The results indicate that bot use in electronic reverse auctions in 2017 represented a more than 5% advantage in successful bid items observed for only 1.99% of the sample bidders, indicated as suspected use. The most relevant indicator for classifying bidders as suspects was the high number of bids issued in relation to the behavior observed in the sample. Results are expected to foster discussion of the effects of bot use on e-trading and to highlight the need for open data policy development for data mining to be an increasingly effective means to assess anomalies and increase the integrity of the bids made by the Federal Government Procurement Portal.