Bayesian Rule Modeling for Interpretable Mortality Classification of COVID-19 Patients

Coronavirus disease 2019 (COVID-19) has been termed a “Pandemic Disease” that infected many people and caused deaths on nearly unprecedented level. As more are each day, it continues to pose serious threat humanity worldwide. result, healthcare systems around the world facing shortage of medical space such as wards sickbeds. In most cases, healthy experience tolerable symptoms if they infected. However, in other patients may suffer severe require treatment an intensive care unit. Thus, hospitals should select who have high risk death treat them first. To solve this problem, number models developed for mortality prediction. lack interpretability generalization. prepare model addresses these issues, we proposed COVID-19 prediction could provide new insights. We identified blood factors affect mortality. particular, focused dependency reduction using partial correlation mutual information. Next, used Class-Attribute Interdependency Maximization (CAIM) algorithm bin continuous values. Then, Jensen Shannon Divergence (JSD) Bayesian posterior probability create less redundant accurate rules. provided ruleset with its own result. The extracted rules form “if antecedent then results, posterior probability()”. If sample matches rules, then result is positive. average AUC Score was 96.77% validation dataset F1-score 92.8% test data. Compared results previous studies, shows good performance terms classification performance, generalization, interpretability.