Feature Analysis and Machine Learning Techniques to determine severity
of COVID-19 infections
Abstract
In 2019 appeared a new Coronavirus Disease (COVID-19) in China,
spreading rapidly globally and causing a pandemic with high infection
and death numbers. To prevent a collapse of the health institutions,
accurate decision making about assignments of intense care units (ICU)
is required, depending on the probable outcome. The usage of machine
learning (ML) for other medical fields had been successful before. So we
applied ML techniques to a dataset of COVID-19 and influenza patients
from Mexico to predict the severity of an individual’s infection
regarding risk factors including, but not limited to, chronic
obstructive pulmonary disease (COPD), cardiovascular disease, diabetes,
asthma, immunosupression, and obesity. We conducted two experiments, one
on hospitalised patients and the other one on a balanced dataset. The
resulting applications should not be used as a diagnostic tool yet, due
to a relatively short time period of data collection and 74.64%
accuracy for the first experiment and 82.61% accuracy for the second
one. Nonetheless it is a good starting point to continue research about
predicting COVID-19 infection’s outcome based on risk factors.