Um método de detecção de outliers para encontrar fraudes na cota para exercício da atividade parlamentar

Abstract

This monograph aims to defend the opening of data as a way to fight corruption, create a shiny application that allows the monitoring of expenses of federal deputies with the Quota to Exercise Parliamentary Activity (CEAP, in Portuguese), develop an unsupervised outlier detection method based on the Kolmogorov-Smirnov test to apply it to the CEAP data set and, using the Monte Carlo method, evaluate the test performance by estimating the probabilities of type I and II errors. We were able to see how an international data opening treaty has been able to inhibit the action of malicious politicians by making their spending on CEAP accessible to any citizen. Simulation studies suggest that as the number of requests a deputy made in the same company increases, the probability the method will detect a small deviation in the distribution of expenses increases as well. When applying the tests to the expenses of a congressperson who was known to defraud CEAP, the method has signaled a set of suspicious companies and among them was the company in which the congressperson committed the fraud.