Classification of Delphi outputs through robust ranking and fuzzy clustering for Delphi-based scenarios

This paper proposes a method for generating robust ranks of Delphi projections, which are particularly suitable as input for clustering algorithms. The resulting clusters can be used for the construction of Delphi-based scenarios. The method is very flexible and can be applied to the classification of any variable derived from subjective judgments. In the analysis and interpretation of the results of a Delphi, a series of problems emerge related to the use of the concept of distance. The use of robust ranks allows us to overcome these problems. The proposed method is also robust with respect to the expertise of panel members, which is a feature that creates many problems both at the moment of measurement and in the subsequent use of those measurements. This opens up important reflections on a crucial aspect of any Delphi study: the dependence of the results on the expertise of the panelists. One of the outputs of the method proposed here is constituted by the uncertainty intervals, which can be used as a monitoring system for the quality of the Delphi projections. By applying the method to the future of families in the north-east of Italy, we will show its validity, reproducibility, and practicality.