COVID-19 Risk Prediction for Diabetic Patients Using Fuzzy Inference System and Machine Learning Approaches. (1st April 2022)

• Record Type: Journal Article
• Authors: Aggarwal, Alok; Chakradar, Madam; Bhatia, Manpreet Singh; Kumar, Manoj; Stephan, Thompson; Gupta, Sachin Kumar; Alsamhi, S. H.; AL-Dois, Hatem
• Other Names: Hamdan Hani Academic Editor.
• Is Part Of: Journal of healthcare engineering. Volume 2022(2022)
• Publication Date: 2022-04-01
• Subjects: Hospital buildings -- Environmental engineering -- Periodicals; Medical technology -- Periodicals; Medical informatics -- Periodicals; 610.28
• DOI: 10.1155/2022/4096950
• Languages: English
• ISSNs: 2040-2295

Abstract

Abstract : Individuals with pre-existing diabetes seem to be vulnerable to the COVID-19 due to changes in blood sugar levels and diabetes complications. As observed globally, around 20–50% of individuals affected by coronavirus had diabetes. However, there is no recent finding that diabetic patients are more prone to contract COVID-19 than nondiabetic patients. However, a few recent findings have observed that it could be at least twice as likely to die from complications of diabetes. Considering the multifold mortality rate of COVID-19 in diabetic patients, this study proposes a COVID-19 risk prediction model for diabetic patients using a fuzzy inference system and machine learning approaches. This study aimed to estimate the risk level of COVID-19 in diabetic patients without a medical practitioner's advice for timely action and overcoming the multifold mortality rate of COVID-19 in diabetic patients. The proposed model takes eight input parameters, which were found as the most influential symptoms in diabetic patients. With the help of the various state-of-the-art machine learning techniques, fifteen models were built over the rule base. CatBoost classifier gives the best accuracy, recall, precision, F1 score, and kappa score. After hyper-parameter optimization, CatBoost classifier showed 76% accuracy and improvements in the recall, precision, F1 score, and kappa score, followed by logistic regression and XGBoost with 75.1% and 74.7% accuracy. Stratified k-fold cross-validation is used for validation purposes.