Clusters of COVID-19 Indicators in India: Characterization, Correspondence and Change Analysis.

We conduct a long-term epidemiology study of COVID-19 in India from Mar 2020 to May 2021 using a number of indicators such as active cases, daily new cases, and deaths, on a micro (district level, per capita) and macro level (state level). Our automated shape-based cluster discovery of the per capita daily new cases (case rate) during the first wave in India (between Mar 2020 and Jan 2021) revealed four distinct shape patterns: sharp-rise and decline, steady-rise and decline, plateau and multiple relatively high peaks. These clusters exhibit a strong geographical correlation. To determine the correspondence between clusters obtained by different indicators, we design a novel metric for determining edge-weights in their intersection graph. This is used for comparative analysis and to develop informative hierarchical cartographic visualizations. We then perform dynamic cluster analysis for different time windows to answer some pertinent questions. Is the second wave similar to or different from the first wave? How has the relative ranking (on micro- and macro-level indicators) of the states varied over the last one year? How much medical resources have been stressed during the peak? We demonstrate that using multiple indicators, we can assess the impact of the epidemic holistically in a particular geography. Our analysis techniques and insights obtained can help the local and state governments in monitoring and managing COVID-19 situation and fine-tuning the ongoing vaccination drive in India.