Classificação automatizada de denúncias no MPBA usando machine learning

Abstract

This article proposes the classification automated reporting area carried out by citizens to the State Public Prosecutor's Office of Bahia (MPBA) through the website “Atendimento ao Citizen”, using learning algorithms machine (machine learning). The process of KDD in the textual bases of complaints aiming to reduce the error rate of the classification carried out by the citizen, currently estimated at 29.43%. With this, it is intended improve efficiency in serving the population, as well as such as, provide savings on institutional resources. Nine classification algorithms were used trained with the complaints database and results were tabulated considering F1 Score and accuracy as main evaluation metrics. With based on the results, there was a reduction in the error rate for 25.00%, highlighting the relevance of this study, in addition to point out some suggestions for improvements to the website and internal administrative procedures of the MPBA.