No abstract

The Office of the Comptroller General (CGU), as the central agency of Brazil's Federal Government Internal Control is responsible for the fiscalization and auditing to fight and prevent corruption. However, some activities such as government purchasing fraud detection are limited by the difficulty of finding effective solutions, considering the huge volume of data, with millions of finantial registers. In such a context, the proccess of knowledge discovery may take advantage of Data Mining techniques, including classification, clusterization and association rules; which associated to multiagent system enrich the processing power through the interation and distribuiton of data mining agents. Thus, this research work used data mining agents with association rules and clusterization techniques to identify cartels, acting in fraud detection. As a research result, more than one hundred association rules were discovered, of which ten have strong evidence of cartelization, proving the usefulness of the approach to support the work of government auditing.

No abstract