Abstract

The COVID-19 pandemic has impacted countries around the world to varying degrees. With a model to capture and compare multiple COVID-related data points simultaneously, the relative success of a country’s plan to contain the COVID-19 pandemic can be determined. Learning to Rank (LTR), the application of machine learning in ranking problems, works through an iterative process to estimate the weight of different data points. By inputting training data with corresponding scores, a LTR algorithm adjusts the value of the data points, thereby creating a model capable of ranking. Using LambdaMART, one such LTR algorithm, three models that ranked countries based on their COVID-19 responses were constructed. The first model utilized data from the United Nations-based Sustainable Development Solutions Network (SDSN) and scored training data using a COVID-19 index that the organization had developed. The second and third models both used data from the Johns Hopkins Coronavirus Resource Center but scored the training data differently, applying either mortality per capita or the COVID-19 index created by the SDSN, respectively. Countries in East Asia and Africa tended to rank higher in COVID-19 response success than those in the Americas and Europe. Urbanized and densely populated countries rated lower than rural countries. In general, countries instituting lockdown protocols or mask-wearing mandates ranked very high, whereas countries resisting these practices placed lower. The data from this study can be used to explore the positive or negative effects of specific policy differences by comparing the ranks of countries with otherwise similar strategies to fight COVID-19.