Abstract

This essay tries to develop the multidimensional perspective to the investigation of social exclusion using a new version of clustering procedures, based on neural networks techniques. Our central aim is to make a fuzzy map of the Italian Well-Being compressing a certain number of indicators as regards as basic lifestyle, subjective well-being, health conditions, social relations, housing problems, durable lack onto a two-dimensional fuzzy space. More specifically we use the Kohonen Map, an unsupervised clustering procedure, to realise a topological segmentation of Italian families onto a two dimensional display along which it is possible to identify differential groups of families featured by different standards of living. The fuzzy Map is made of prototypical cluster each of which recognise families (and individuals) featured by a particular combination of deprivation and well-being indicators. Families (or individuals) situated in a closeness tend to share similar characteristics, while families (farther) in the fuzzy map show pronounced differences. We make use of the Multiscopo ISTAT dataset (2001) to explore the structure of correlations among the indicators and to recognise different forms and level of vulnerability and social exclusion. With multilevel logistic regression models we will demonstrate how education, employment status and life events continue to be powerful predictors of the multiple deprivation and of the social exclusion in Italy although with significant variations across geographical regions.