An algorithmic pipeline based on unsupervised machine learning algorithms, which aims to operate in tandem with physicians and provide additional knowledge for the proper categorization of COVID-19 infected patients based on their severity, is proposed in this study. Data from patients hospitalized in our clinic are collected and stored in separate Microsoft Excel files (.xlsx), which are loaded into memory. A script is concatenating them all into a single dataframe where they are checked for NaN (Not a Number) values. Because of the nature of the data, patients with missing information are discarded entirely from the dataset, since information inference would be a biased practice for the particular application. Next, we apply data normalization by scaling all numerical variables between the (0,1) range, so that the range of all numerical variables is the same, and any bias towards certain variables is avoided .A thorough and detailed data collection process was designed in order to collect information for the patients, without disturbing the clinical treatment, or upsetting them in the process.
