About the Project
RiskGuard: Effective Risk Assessment in Public Procurement
Coordinated by NOVA IMS, the RiskGuard project addresses the critical challenge of improving transparency and accountability in public procurement. As public institutions increasingly manage large and complex datasets, there is a growing need for intelligent systems that support risk identification and audit decision-making.
RiskGuard pioneers the development of a transparent, machine learning-based decision-support system to detect risks and irregularities in procurement data. Grounded in data science, econometrics, and AI, the initiative aims to equip the Portuguese Court of Auditors with innovative tools to modernize its oversight capabilities.
The project leverages supervised and unsupervised machine learning, network science, and statistical modeling to identify patterns of potential corruption, including single bidding, collusion, and contract anomalies. It directly addresses calls from European regulatory bodies to embed digital intelligence into public sector auditing practices.
Through three structured phases—data integration and analysis, algorithmic model development, and deployment and validation—RiskGuard delivers a scalable and explainable solution that strengthens institutional resilience against procurement-related fraud.
As part of its implementation, the project will:
-
Construct a data lake integrating procurement data from national institutions such as the Court of Auditors, Tax Authority, and Institute of Registries and Notary.
-
Develop models using methods like Random Forests, Neural Networks, SVMs, and logistic regression, supported by network analysis where applicable.
-
Validate models using cross-validation techniques and standard performance metrics to ensure robustness.
-
Deploy a user-friendly dashboard that allows auditors to explore indicators and interact with risk models in real-time.
-
Provide comprehensive documentation, training, and technical support to ensure long-term adoption.
-
Share findings through international publications, conference presentations, and a final public Policy Impact Workshop.
