Work

Anomaly Detection in Portuguese Public Procurement Contracts: An Embedding-Based Machine Learning Approach

Year: 2025
Type: dissertation
Abstract: Public procurement is an essential aspect of public administration and one of the functions that help fulfill the population's needs, although it is prone to irregularities. With the evolution of digital infrastructures, many governments have adopted electronic procurement systems that automate workflows, improve transparency, and increase access to contract information. These advancements make it possible to employ techniques that analyze large volumes of procurement data to uncover unusual patterns and behaviors. Considering that, this study uses real-world data extracted from the Portuguese public procurement platform BASE, covering the years 2020 to 2024, to explore the application of unsupervised machine learning algorithms, particularly clustering, for anomaly detection. A key innovation in this research is the creation of text embeddings, a technique rarely applied in this field. The resulting model aims to enhance fair competition and good governance by identifying deviations from normative procurement processes. Additionally, it provides auditors and policymakers with the tools to effectively identify potential irregularities.
Author: Silva, Ana Rita Saraiva da
Language: English
Open Access status: green
OpenAlex ID: W7111848468
Cites: 71
Cited by: 1
Related to: 10