Robust interactive and interpretable hierarchical PCA: Applications to crime data

In a context characterized by increasing institutional complexity and growing constraints on available resources, this paper proposes an interpretable framework for the construction of composite indicators aimed at supporting strategic decision making in public administration with specific emphasis on financial crime data. Robust statistical methods provide valuable support for analysing how and where resources should be deployed, enabling data-driven decisions aimed at improving the performance of Public Administrations for the benefit of the community. However, such analyses must produce results that are interpretable, clearly quantify the contribution of each unit, and remain robust in the presence of atypical observations. Furthermore, the inclusion of interactive tools is essential to allow end users to easily explore the characteristics and relative positions of units within the analytical space. Motivated by these requirements, this paper describes the interactive and interpretable framework based on robust hierarchical principal component analysis for optimizing resources and operational strategies within the Guardia di Finanza (GdF - Italian Economic and Financial Police).