Implementation of a self-tuning algorithm for Spectral Clustering. Applications on earthquakes and COVID-19

This thesis aims to present the self-tuning algorithm implemented to perform spectral clustering. After having discussed the motivations that lead to the introduction of this clustering method, its way of operating is explained in detail. The innovation of the constructed algorithm consists in the ability to automatically select a suitable default value for each of its parameters, with the latter representing the viable alternatives in the application of the spectral clustering. First, the algorithm is applied on a dataset concerning the occurrence of earthquakes around the world. Next, the application on data regarding the spread of the COVID-19 virus in the U.S. counties is discussed.