Graph Data Mining for Detecting Collusions in Bidding Processes: A Case Study

Bidding collusion causes significant financial damage in Brazil and compromises the efficiency of public service delivery. This work proposes the representation of bidding disputes as a directed graph and the use of data mining techniques on graphs to detect suspicious data patterns that may reveal anti-competitive behaviour. The algorithms and techniques used were tested in a case study with the real bidding data of a Brazilian mixed economy company. The results show that to ensure the integrity of the bidding processes it is required to develop more complex preventive controls which check more comprehensively the entire network of relationships in the graph.