Machine Learning Based Approach to Assess Territorial Marginality

Abstract

AbstractThe territorial cohesion is one of the primary objectives for the European Union and it affects economic recovery pushing the role of Public Administration in promoting territorial development actions. The National Strategy for Inner Areas (SNAI) is a public policy promoting endogenous development processes in marginal territories with low settlement density. Specific contexts where rules and standards defined for the organization of large metropolitan aggregates lose their effectiveness whose identification represents a critical stage for policy efficacy and the actual map of SNAI target areas appears to be the results of a weak and simplified analytical approach. These considerations are the origin of the research question that underlies this work: identify the typical characteristics of Basilicata's marginal areas through machine learning techniques and, subsequently, reclassify the national territory using the trained model. However, outlining the boundary of this territories is only a preliminary task. The following step is the identification of the dynamics within the different territorial sub-systems that make up the inner peripheries. This paper presents the results of the local model-agnostic method for interpreting the obtained results. It emerges, thought cooperative game theory by Shapley values, the need to refine analytical methods that are sensitive to the measurement of the different context conditions. Future perspectives of the research regard the extensive deepening of the application on the basis of wider datasets able to make explicit spatial components of the distribution of the observed phenomena.KeywordsTerritorial cohesionInner peripheryInterpretable machine learning