dynamic inventory risk profiling using pca and clustering: a data-driven approach to supply chain optimization

Effective inventory management is critical for mitigating inefficiencies such as overproduction, excessive holding costs, and stockouts. this study leverages dbscan and gmm clustering methods, combined with principal component analysis (pca) for dimensionality reduction, to categorize inventory data into distinct risk-based clusters. the analysis highlights that dbscan outperformed gmm, achieving a silhouette score of 0.62 compared to 0.49, while identifying three meaningful inventory clusters. each cluster reflects unique combinations of risk factors, providing actionable insights for optimizing inventory levels. the study demonstrates how these clusters enable targeted strategies to address inefficiencies and improve overall inventory management. limitations include the reliance on historical data, which may not fully capture dynamic market conditions, and the assumption of fixed clustering parameters. the findings underscore the importance of choosing clustering algorithms suited to the data’s characteristics and highlight the potential of pca in enhancing computational efficiency. future research should explore dynamic clustering techniques and integrate real-time data streams to refine inventory management strategies further.