COVID-19 Patient Count Prediction Using LSTM.

Area of Science:

• Epidemiology
• Computational Biology
• Machine Learning

Background:

• The COVID-19 pandemic necessitated accurate forecasting of patient volumes for effective public health response.
• Exponential spread patterns of COVID-19 required computational modeling for resource allocation and hospital load management.
• Accurate patient volume estimation is crucial for governmental decision-making during pandemics.

Purpose of the Study:

• To predict the volume of COVID-19 patients in Pakistan using a Long Short-Term Memory (LSTM) neural network.
• To evaluate the effectiveness of the LSTM model in forecasting COVID-19 patient numbers.
• To compare the LSTM model's predictions with existing models for accuracy.

Main Methods:

• Utilized a Long Short-Term Memory (LSTM) recurrent neural network (RNN) for time-series prediction.
• Trained the RNN model on COVID-19 data from Pakistan spanning March to May 2020.
• Predicted the percentage of positive COVID-19 patients for June 2020 and calculated Mean Absolute Percentage Error (MAPE).

Main Results:

• The LSTM model demonstrated effectiveness in predicting COVID-19 patient volumes in Pakistan.
• Model performance was analyzed across various LSTM units, batch sizes, and epochs.
• Predicted patient counts closely matched actual patient data, outperforming a comparative prediction model.

Conclusions:

• LSTM models offer a viable approach for forecasting infectious disease outbreaks like COVID-19.
• Accurate patient volume prediction supports informed public health strategies and resource management.
• The developed LSTM model provides a reliable tool for estimating future COVID-19 patient loads in Pakistan.