Fuzzy Deep Neural Network Technique for Prediction of Morality Risk in Patients with Corona Virus Disease

Predicting risks in patients with Corona Virus Disease 19 (COVID-19) involves assessing the likelihood and severity of various outcomes, such as hospitalization, ICU admission, disease progression, or mortality. Machine learning models, including deep neural networks, can be used for this task by leveraging patient data to make predictions. In contrast to using deep learning techniques for COVID-19 disease detection through image analysis, an approach was proposed that combined feature selection, instance clustering and Deep Neural Network (DNN) to forecast the threat of morality in COVID-19 patients. However, DNN struggles with handling uncertain and imprecise data. In addition to this, DNN is sensitive to variations and outliers in the input data. In this paper, Fuzzy DNN (FDNN) is introduced to handle the uncertainty, data variation and outliers in input data for better prediction of risks in patients with COVID-19. FDNN is a deep neural network that combines fuzzy and neural representations to derive information. The knowledge obtained from both views is integrated to create the final data representation for prediction of risks in patients with COID-19. The findings indicate that the suggested approach using FDNN yields superior accuracy, precision, and recall in predicting COVID19 patients.