Identification of Governance Risks in a Campus Environment Using Clustering Algorithms: A Comparative Study Based on Data Mining

Abstract

Governance risks in higher education remain a critical challenge, influencing both performance and accountability. Despite increasing attention to risk analysis, the application of data-driven approaches—particularly clustering algorithms—for governance risk identification is still limited. This study proposes and evaluates a web-based information system integrating clustering algorithms to classify and visualize non-academic governance risks. Using the CRISP-DM framework, 678 risk records from the Internal Audit Unit (SPI) of Politeknik Negeri Malang were analyzed. The system employed K-Means clustering on three attributes: performance indicators, budget allocation, and risk levels. Additional modules supported data upload, automated preprocessing, historical logging, and visualization. Comparative evaluation with DBSCAN and Hierarchical Clustering was conducted using Silhouette Scores. Results indicated that K-Means achieved the highest score (0.48), outperforming DBSCAN (-0.705) and Hierarchical Clustering (-0.395). K-Means produced five interpretable clusters representing very high, high, medium, low, and very low risk priorities. Functional and usability tests confirmed the system’s reliability and user-friendliness in supporting automated governance risk assessment. Findings demonstrate the effectiveness of K-Means clustering in segmenting governance risks and providing actionable insights for audit teams. Future work should integrate predictive models and real-time data to strengthen proactive governance risk mitigation.