Machine Learning Applied to Open Government Data for the Detection of Improprieties in the Application of Public Resources

Context: Making government data publicly available is an important mechanism of transparency and social control. In this regard, inumerous laws have made it mandatory to divulgate government procurement data. Problem: The large volume of unstructured textual information available on government portals is an obstacle to effective social control. Making it difficult to do more in-depth analyzes of public spending. Solution: Use of Machine Learning algorithms to perform text mining and grouping items acquired by public administration. Labeling public purchases and grouping similar items, in order to facilitate the detection of improprieties in government purchases. IS Theory: This work is associated with the Theory of Computational Learning, which aims to understand the fundamental principles of learning and design better-automated methods. Method: The article is a case study, and its evaluation was executed with the support of specialists in the field. The results were analyzed based on a quantitative approach. Summary of Results: The results observed in the evaluated cases were promising, the resulting clusters from the application of the solution had sufficiently coherent semantic values, in order to allow more complex analyzes of government purchases. Contributions and Impact in the IS area: The results show that applying text mining and machine learning techniques can extract useful information from government purchases data and allowing to perform better analyzes of public spending.