A COVID-19 forecasting system for hospital needs using ANFIS and LSTM models: A graphical user interface unit.

Area of Science:

• Medical Informatics
• Artificial Intelligence
• Epidemiology

Background:

• Centers for Disease Control and Prevention data indicates a significant hospitalization rate for COVID-19 patients with underlying conditions.
• Approximately 33% of hospitalized COVID-19 patients required intensive care unit (ICU) admission.
• Accurate prediction of COVID-19 hospital admissions is crucial for resource management.

Purpose of the Study:

• To identify a machine learning algorithm capable of accurately predicting COVID-19 hospital admissions.
• To develop a predictive model for forecasting the number of intensive care unit (ICU) and non-ICU COVID-19 patients.
• To create a user-friendly interface for real-time COVID-19 case prediction.

Main Methods:

• Utilized daily COVID-19 case data from July 2020 to November 2021.
• Trained Long Short-Term Memory (LSTM) network, Adaptive Neuro-Fuzzy Inference System (ANFIS), Support Vector Regression (SVR), and decision tree models.
• Validated models using a 60-day test dataset to predict ICU and non-ICU patient numbers.

Main Results:

• All tested models successfully predicted COVID-19 case dynamics in ICU and non-ICU wards.
• The ANFIS model exhibited superior predictive accuracy compared to LSTM, SVR, and decision tree models.
• Model performance was assessed using root-mean-square error and R-squared metrics.

Conclusions:

• AI-based forecasting models, including ANFIS, LSTM, SVR, and tree regression, are vital for managing healthcare resources during pandemics.
• The developed graphical user interface facilitates the prediction of COVID-19 cases for up to one week ahead.
• The study provides a tool for optimizing healthcare resource allocation amidst the COVID-19 pandemic.