A Machine Learning Approach on the Problem of Corruption

Abstract

This work presents a step-by-step building of a model that effectively classifies a given municipality as corrupt or not. The output is the likelihood of the city being corrupt, which can be a valuable tool in preventing future corruption cases. This model was constructed to utilize the already existing API from the São Paulo State Court of Auditors and was built to deploy monthly reports. The XG Boosting model was the most robust among the many models trained and presented the best AUC score and accuracy.