Detection of Anomalous Proposals in Governmental Bidding Processes: A Machine Learning-Based Approach

Abstract

Government procurement involves a formal process wherein government bodies select supplier proposals for goods and services to obtain the best possible terms. This study employs three machine learning algorithms to detect irregularities in the Brazilian government’s procurement processes, focusing on data from Paraiba state. The efficacy of these algorithms was evaluated using a controlled dataset that contains known anomalies, assessing their ability to identify deviations. The findings demonstrate the effectiveness of these methods, notably the One-Class SVM, which excels at revealing patterns indicative of possible irregularities in government procurement. In conclusion, this research underscores the potential of machine learning algorithms in enhancing the transparency and integrity of public bidding processes.