Predição do absenteísmo em agentes de segurança pública usando aprendizagem profunda

Detalhes bibliográficos
Ano de defesa: 2019
Autor(a) principal: Silva Júnior, Edival Lima da
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal de Alagoas
Brasil
Programa de Pós-Graduação em Informática
UFAL
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: http://www.repositorio.ufal.br/handle/riufal/5889
Resumo: Absenteeism is a complex phenomenon that is expressed by the physical absence of the individual, usually at his job. In public security institutions, these absences bring many personal, social and economic losses, in addition to occurring in percentages superior to those of the other professional categories. Thus, determining its preponderant factors and allowing preventive actions to be carried out effectively would bring numerous benefits to these institutions and their agents. In this work, we propose and evaluate predictors capable of identifying the most prone agents to long-term absenteeism. These predictors should make decisions based on the professional history of each agent extracted from databases of public security institutions. We carried out experiments using a database comprised of 6 years of absences from agents of the Military Police of Alagoas in Brasil, from which we selected attributes that would be correlated to the phenomenon of absenteeism.We evaluated Deep Learning architectures such as Multilayer Perceptron, Long Short-term Memory and Recurrent Neural Network. We applied attributes selection techniques and compared the results obtained by the Machine Learning Support-Vector Machine technique.We present the best architectures for predicting the prolonged absenteeism of agents, which proves that it is possible to predict this type of absence, reaching 78% of accuracy, which would support the implementation of effective prevention measures in these institutions. Finally, we conclude that the use of data referring to a more significant number of years results in better results in the prediction of absenteeism.