Aplicação de Rede Neural Colaborativa à Classificação de Consumidores de Serviços Públicos

• Ano de defesa: 2021
• Autor(a) principal: JESUS, Lucas Gabriel Rezende de
• Orientador(a): TEIXEIRA, Mário Antonio Meireles
• Banca de defesa: TEIXEIRA, Mário Antonio Meireles , ALMEIDA NETO, Areolino de , CARMONA CORTÊS, Omar Andrés , ISHII, Renato Porfirio
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Maranhão
• Programa de Pós-Graduação: PROGRAMA DE PÓS-GRADUAÇÃO EM CIÊNCIA DA COMPUTAÇÃO/CCET
• Departamento: DEPARTAMENTO DE INFORMÁTICA/CCET
• País: Brasil
• Palavras-chave em Português: concessionárias de serviço público; classificação de risco; CollabNet; aprendizagem profunda
• Palavras-chave em Inglês: public service concessionaires; credit scoring; CollabNet; deep learning
• Área do conhecimento CNPq: Sistemas de Computação
• Link de acesso: https://tedebc.ufma.br/jspui/handle/tede/3989

In the 1920s the brazilian state instituted a political and economic decentralization process, one of the main consequences of which was a concession to private companies of the right to economically develop and explore a public service, until then was the sole and exclusive responsibility of the State. Since then, these concessionaires have search to balance the provision of equitable and quality public service with the collection process inherent to every company. In order to offer parameters that bring more objectivity to this process, in this work we propose a methodology for classifying the default risk profile of consumers. The use of the deep collaborative neural network, CollabNet is presented in a customer database of a utility company. The methodology presents promising results such as an accuracy of 88.1 %, a sensitivity of 93.9 % and a negative predictive value of 93.1 %. It is still suggested an incorporation of new characteristics about consumers such as geographic aspects and family income in order to improve the results obtained.