Unsupervised Clustering in Epidemiological Factor Analysis

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