Ciências de dados nos controles internos dos entes e empresas públicas

Abstract

Current computing capacity has enabled the constant development of tools to support the continuous improvement of processes and internal controls in public entities and companies, in which Data Sciences should be highlighted. Given this, the present work aimed to test the applicability of Data Science techniques to public domain data from public entities, looking for relevant anomalies in the records. A model was built to identify suspicious data using techniques based on Benford’s Law, k-NN and DBscan (Outlier Analysis), promoting satisfactory and promising results. It was possible to verify that the application of Data Science techniques is still incipient in some layers of controls in public bodies, which opens a window of opportunity for the implementation of procedures linked to the construction of models for the purpose of population verification of documentation and database to mitigate fraud events and illegal acts committed by public agents and public administration service providers. In practical terms, the insertion of Data Science techniques into the internal control procedures of public entities and companies depends on sponsorship and investments from managers, to provide an increase in computational capacity and training of employees, and mainly, breaking the systems paradigm of conventional controls based on documentary verification by sampling.