Multicriteria Anomaly Detection in Government Purchases

Abstract

Government purchases are the usual instrument for public acquisition of goods and services. Despite extensive legislation and several control and auditing mechanisms, frauds are still diverse and commonplace at all levels of public administration. This work proposes a methodology for detecting anomalies in government purchases. The methodology promotes several levels of filtering with respect to entities involved and purchases considered as fraudulent considering diverse criteria. The applicability and effectiveness of the methodology is demonstrated through a case study using real data where we were able to identify a long term provider collusion.