Clustering COVID-19 Mortality Time Series

The undisputed assumption during the COVID-19 pandemic, was that the isolated pathogen SARS-COV-2 was spreading worldwide with the same mortal effects. However, examining the mortality records it is possible to appreciate substantial differences between countries in their response to the pandemic. This paper looks for clusters of countries that may shate some features explaining the diversity of the response. First, the paper extracts latent mortality patterns by Principal Component Anaylsis (PCA), Independent Component Analysis (ICA), and Non-negative Matrix Factorization (NNMF). Clustering is then carried out over the coefficients of the respective transformations of the countries mortality time series. The choice of the number of latent patterns has been guided by the explained variance in PCA and the avoidance of overfitting taking into account the ratio of representatives to the size of the dataset. Additionally, direct clustering of the mortality time series has been carried out by K means. This paper discusses some common findings that point to cultural and socio-economic factors underlying the response.