Variable Selection for Meaningful Clustering of Multitopic Territorial Data

This paper proposes a new methodology to improve territorial cohesion in clustering processes where many variables from different topics are considered. Clustering techniques provide added value identify typologies, but there still unsolved challenges when data contain an unbalanced number of topics. The feature selection method (TFSM) is presented as select the representative variable each topic such that interpretability resulting clusters preserved and geographical improved with respect classical approaches. also introduces thermometer knowledge acquisition tool allows experts transfer semantics mining process. TFSM index potential explainability (Ek) criteria most promising for clustering. Ek based on combination inferential testing metrics support. proposal applied INSESS-COVID19 database, groups vulnerable populations were found. A set 195 21 thematic blocks used compare results traditional multiview analysis both point view capacity support further decision making.