Novel Cluster Analytics for Analyzing COVID-19 Patient Medical Data

Abstract

The ongoing global COVID-19 pandemic has caused more than 440,000 deaths among more than 8 million cases globally by Mid-June, 2020. This pandemic has caused a staggering worldwide socioeconomic impact and loss of lives. This research proposes an innovative technological approach to analyze COVID-19 patient data for new analytical insights via developing a transformative pattern identification algorithm for cluster analysis in tabular numerical data tables, e.g., patient medical data files and in disease networks. The underlying mathematics is based upon Lie algebras and continuous Markov transformations that are foundational in quantum theory, relativity, and theoretical physics. Our novel algorithm does not use an arbitrary concept of proximity or nearness but instead is based upon an information flow model where clusters are identified, and rank ordered by the matrix eigenvalues. The component clusters identify the degree of patient cluster participation by the nodal weight given by each associated eigenvector. Medical metadata tags in the tables are automatically linked to the cluster eigenvalue and eigenvectors to facilitate interpretation of the analytics. The core algorithm has been coded and will be ported to a cloud environment allowing other investigators to submit data files for cluster analytics. We plan to analyze COVID-19 patterns and expect to work with other medical research teams on pattern identification in deidentified medical patient data sets. We expect this ongoing research to lead to significant practical and theoretical insights and a greater understanding of our transformative network clustering algorithm at the individual COVID-19 patient level, hospital level and beyond.