Recurrence as a Governance Signal: Diagnostic Network Metrics for Public Procurement Oversight in Greece

This record contains the processed dataset and full set of empirical results supporting the study: Fountoukidis, I.G., Dafli, E.L., Antoniou, I.E., and Varsakelis, N.C. (2026). "Recurrence as a Governance Signal: Diagnostic Network Metrics for Public Procurement Oversight in Greece." The dataset includes: - The final feature matrix used for model training (buyer–supplier pairs) - Network-based predictors, including: - Historical Frequency (HF) - Preferential Attachment (PA) - Adamic–Adar (AA) - Model performance metrics across experiments: - AUC, F1-score, Precision, Recall - Results across different negative sampling ratios - CPV-level analyses and temporal splits (train year t, test year t+1) - Summary statistics of positive and negative class distributions The repository enables full reproduction of the modelling and evaluation stages presented in the paper. Note that the following components are not included: - Raw KIMDIS procurement data - Full graph database construction (Neo4j pipeline) This dataset corresponds to the final analytical outputs used in the published results.