Aplicação de machine learning para classificação do grau de confiança de servidores públicos federais em gestores e demais agentes públicos

Abstract

Data mining and machine learning are tools that are increasingly used nowadays for data analysis in the most diverse sectors of society. Corruption in the public service is the subject of several studies that seek to understand this phenomenon, however, there is still a lack of research that analyzes the impacts of this type of conduct at the organizational level from the application of machine learning. For this reason, machine learning techniques are presented as a tool to study this scenario. In this way, the goal of this research was to apply data mining algorithms to classify the level of confidence of Brazilian federal public servants in the leaders and other public agents of their organization. For that, a public database was used with answers to a questionnaire on ethics and corruption in the federal public service. The results indicated that the best accuracy obtained was 87% for predicting confidence in managers and 85% for coworkers. In addition, the study identified that the main factors considered by the algorithms for ranking trust were the observation of cases of corruption, unethical conduct, undue favoritism, and transparency inside the organization. In conclusion, the predictive model clarified relevant aspects in the perception of trust within public organizations.