Application of nonparametric pattern recognition algorithms in the characterization and forecasting of geophysical events

Abstract

In this Ph.D. thesis, I use a multivariate statistical approach to characterize some relevant Geophysical processes. The empirical methods I have chosen are multivariate analysis techniques belonging to the class of Pattern Recognition algorithms. The potentiality of this type of analysis is due to its ability in identifying possible repetitive schemes (patterns) among objects belonging to distinct categories, by extracting information from any possible combination (linear or not) of variables that are supposed to have an influence on the process.