POVERTY ANALYSIS WITH NEURAL NETWORKS -Italy case study-

Abstract

The aim of this dissertation is to propose a new approach to poverty multidimensionality analysis. In particular, the research question concerns a comparison between the classical statistical models of probit/logit regressions and neural networks. The empirical analysis is done according to European Union’s statements and variables declaration as a reference benchmark for Italy. The thesis is divided into four parts, namely a first introductory section concerning a brief literature review about poverty according to the European and Italian contexts; the second, regarding the proposed methodologies of regression models and neural networks theoretical background, followed by a third concerning data description. The core of the elaborated is in the fourth chapter, where the emprirical analysis is carried out: the main results, both derived by the traditional regressions, and the ones provided by the created neural network, will be compared.