Government Procurement Contract Graphs for Repayment Risk Prediction in Industrial Suppliers

Abstract

Government procurement contract financing allows industrial suppliers to obtain working capital based on confirmed purchase orders, delivery contracts, acceptance records, and expected public-sector payments. Although these contracts are usually considered stable, repayment risk may emerge from delayed acceptance, contract modification, weak delivery capability, abnormal subcontracting, disputed invoices, and borrower-side cash-flow pressure. This study proposes a government procurement contract graph model for repayment risk prediction in industrial suppliers. The model links suppliers, government buyers, procurement contracts, bid records, delivery milestones, acceptance certificates, invoices, subcontractors, bank loans, litigation records, and repayment outcomes. A graph neural encoder captures supplier-contract dependency, while knowledge inference identifies risk paths involving repeated contract delay, abnormal subcontracting chains, buyer-side payment postponement, invoice inconsistency, and supplier credit deterioration. Experiments are conducted on a procurement finance dataset containing 33,200 industrial suppliers, 8,460 government buyers, 1.18 million procurement contracts, 2.36 million delivery milestones, 740,000 invoice records, 41,600 financing contracts, and 5,380 confirmed repayment-risk cases over 40 months. Compared with a supplier-level credit scoring baseline, the proposed method shortens median warning time from 73 days to 27 days before repayment deterioration. Graph reasoning identifies 6,920 procurement-related risk paths and 2,310 delayed-acceptance chains. Manual contract review is reduced from 19,800 contract batches to 4,420 graph-linked investigation cases after relation aggregation. The inference engine completes full monthly assessment in 11.6 minutes and processes 3.54 million contract-related edges per update. The results demonstrate that government procurement contract graphs can strengthen repayment risk prediction by connecting contract execution, public-sector payment behavior, subcontracting relations, and supplier financial stress.