Unsupervised Clustering in Epidemiological Factor Analysis

Abstract

The analysis of epidemiological data at an early phase of an epidemiological situation, when the confident correlation of contributing factors to the outcome has not yet been established, may present a challenge for conventional methods of data analysis. This study aimed to develop approaches for the early analysis of epidemiological data that can be effective in the areas with less labeled data. An analysis of a combined dataset of epidemiological statistics of national and subnational jurisdictions, aligned at approximately two months after the first local exposure to COVID-19 with unsupervised machine learning methods, including principal component analysis and deep neural network dimensionality reduction, to identify the principal factors of influence was performed. The approach and methods utilized in the study allow to clearly separate milder background cases from those with the most rapid and aggressive onset of the epidemics. The findings can be used in the evaluation of possible epidemiological scenarios and as an effective modeling approach to identify possible negative epidemiological scenarios and design corrective and preventative measures to avoid the development of epidemiological situations with potentially severe impacts.