HAP-COVID: Hybrid Model Integrating Machine Learning and Neutrosophic Statistics to Estimate Mortality Risk in COVID-19 Comorbid Patients

Abstract

This study aims to utilize a hybrid predictive model to evaluate the mortality risk in COVID-19 patients with comorbidities. It is challenging to make clinical decisions while dealing with SARS-CoV-2 because of the high death rate related to chronic conditions such as diabetes, hypertension, chronic obstructive pulmonary disease (COPD), renal failure, heart disease, and cancer. To overcome this obstacle, a Monte Carlo algorithm-based simulated database was created, with clinical variables represented by Bernoulli and truncated normal distributions. Several prediction models, including Decision Trees, Naive Bayes, and Neutrosophic Statistics, were trained using this database. While the neutrosophic model permitted risk categorization according to truth, falsity, and indeterminacy, the decision tree model outperformed the Naive Bayes model in terms of accuracy. When it came to controlling diagnostic uncertainty, the hybrid approach worked well. Ultimately, this technology provides significant assistance in intricate clinical settings, enhancing the process of medical decision-making.