The assessment of resilience is the core of any governance and risk management analysis for shocks and crises. This assessment is generally dependent on similarity measurement for the identification of patterns and relationships between different individuals or groups of individuals within a given community. One of the difficulties in studying social resilience processes is the lack of appropriate analytical tools that take into account the dimensions of resilience. The use of conventional similarity measures can lead to some bias in the analysis and consequently to errors in decision-making. In this paper, we propose a new measure of similarity for the calculation of the degree of similarity between two individuals described by several univalued and multivalued variables of heterogeneous types (quantitative, qualitative or textual). Our proposal, compared to most of the similarity measures presented in the literature, has the merit of directly exploiting a table of heterogeneous data containing both univalued and multivalued values (intervals, sets, textual, etc.). Generally, a homogeneous transformation of the table is used and then a classical similarity index is used for the construction of the similarity matrix. However, this homogeneity of the table leads to distortions and negatively influences the expressive character of the data. The comparison of our approach with other proposals in the literature according to the Davies-Bouldin and Silhouette quality index gives us the best values for these indices, demonstrating its effectiveness for studying social resilience processes.
