Abstract: The complexity, persistence, and concealment of contamination in groundwater sites pose significant challenges to risk management. To address these challenges, this study develops a knowledge graph-based intelligent decision-making model for managing contaminated groundwater sites and proposes an intelligent risk management framework that integrates technical, economic, and social dimensions. By consolidating a risk management technology repository and a case study database for contaminated groundwater sites, a knowledge graph with semantic reasoning capabilities is constructed. Additionally, the KG-RF (Knowledge Graph - Random Forest) intelligent decision-making model is designed by integrating knowledge graphs with random forest algorithms to accurately recommend risk management solutions. During the model training phase, the KG-RF model was trained on a dataset of contaminated sites recorded in the knowledge graph and achieved 94.86% accuracy on the test set. A case study on a contaminated groundwater site demonstrates that the KG-RF model, based on key indicators such as hydrogeological conditions and pollutant characteristics, identifies and matches similar historical cases, calculates similarity scores, and recommends optimal risk management solutions. The results indicate that the dual-layer PRB (Permeable Reactive Barrier), consisting of ZVI (zero-valent iron), granular activated carbon, and biofilm, achieves the highest similarity score (0.9278). This technology excels in terms of technical maturity, applicability, and cost-effectiveness. The findings indicate that knowledge graphs have significant application potential in risk management for contaminated groundwater sites, contributing to the enhancement of intelligent and precise risk management. Additionally, they provide new methods and approaches for smart decision-making in complex environmental issues.