A Self-Organizing Map of the Elections in Portugal

Abstract

As (artificial) neural networks are simulations of the supposed biological neurons work, the structure of human brains - where processing units, the so-called neurons, are connected by synapses - is approximated by (artificial) neural networks. As most of neural networks, self-organizing maps are trained through a learning process. By the use of a neighborhood function in this learning process, self- organizing maps (SOMs) thus allow to visualize which (and how) democratic elections were more similar/distinct. For Portugal the SOM identifies two clusters of elections: one made of those corresponding to a re-election of the incumbent, i.e. in 1987, 1995, 1999 and 2009; and another made of elections that led to a change in the party in power, i.e. 1991, 2002, 2005 and 2011.