On the Implementation of the Artificial Neural Network Approach for Forecasting Different Healthcare Events.

Area of Science:

• Epidemiology
• Data Science
• Public Health

Background:

• The COVID-19 pandemic in Pakistan has significantly impacted healthcare and other sectors.
• Accurate forecasting of confirmed cases, deaths, and recoveries is crucial for effective policy decisions.

Purpose of the Study:

• To compare the predictive accuracy of an artificial neural network (ANN) model against five univariate time series models for COVID-19 data in Pakistan.
• To evaluate model performance using statistical measures and the Diebold-Mariano test.

Main Methods:

• Utilized daily COVID-19 dataset for Pakistan (March 10 - July 3, 2020) including confirmed cases, deaths, and recoveries.
• Applied and compared an artificial neural network model with five univariate time series models (e.g., Moving Average, Autoregressive Moving Average).
• Assessed model performance using two statistical measures and the Diebold-Mariano test for mean error accuracy.

Main Results:

• The artificial neural network model demonstrated superior performance in predicting COVID-19 deaths and recovered cases.
• The Moving Average model was identified as the best-performing model for forecasting confirmed COVID-19 cases.
• The Autoregressive Moving Average model ranked as the second-best for confirmed case prediction.

Conclusions:

• Artificial neural networks are effective for predicting COVID-19 mortality and recovery trends.
• Simpler time series models like Moving Average are highly effective for short-term confirmed case forecasting.
• Model selection is critical for accurate epidemiological predictions to inform public health strategies.