B0762
Title: Unsupervised machine learning for estimating the socioeconomic vulnerability in the European Union
Authors: Angeles Sanchez - Universidad de Granada (Spain) (Spain) [presenting]
Eduardo Jimenez-Fernandez - Universidad de Granada (Spain)
Abstract: The main aim is to provide an alternative criterion for allocating the structural funds among European Union regions for 2021 to 2027 that better reflects citizens' quality of life. Using the vector space formed by all the observations, the distance learning or DL2 method is applied to build a composite index of socioeconomic vulnerability for the 233 regions of the European Union. More specifically, the DL2 composite indicator represents a weighted Euclidean metric where the weighting scheme is estimated with unsupervised machine learning techniques. This method is based on the mathematical concept of distance or metric, enabling comparisons between the studied units. To develop a system of indicators capable of representing how a region can respond to the pressures and challenges of the Cohesion Policy, 16 single indicators are selected. Eight representative indicators of the socio-economic weakness or fragility of the regions and eight representative indicators of the capacity of the regions to face challenges or structural changes have been chosen. The results show that following the multidimensional approach to allocating the structural funds, there are remarkable differences in the maps of priority regions.