Clustering time series: an application to COVID-19 data

In this paper we present an attempt of clustering time series focusing on Italian data about COVID-19. From the methodological point of view, we first present a review of the most important methods existing in literature for time series clustering. Similarly to cross-sectional clustering, time series clustering moves from the choice of an opportune algorithm to produce clusters. Several algorithms have been developed to carry out time series clustering and the choice of which one is more adapt depends on both the aim of the analysis itself and the typology of data at hand. We apply some of these methods to the data set of daily time series on intensive care and deaths for COVID19 stretching from, respectively, 23/02/2020 to 15/02/2022 and from 23/02/2020 to 29/03/2022. These data refer to the 19 Italian regions and the two autonomous provinces of Trento and Bolzano.