Predictive Modeling for Pandemic Forecasting: A COVID-19 Study in New Zealand and Partner Countries.

Area of Science:

• Epidemiology
• Computational Biology
• Public Health

Background:

• Accurate early-stage prediction of COVID-19 spread is crucial for effective public health interventions.
• Leveraging large-scale datasets and advanced machine learning models can improve forecasting accuracy.
• Understanding geographic variations and temporal trends is key to modeling disease dynamics.

Purpose of the Study:

• To propose and evaluate a data-driven approach for enhancing COVID-19 spread prediction in early stages.
• To systematically compare the performance of three machine learning models: ARIMA, Prophet, and LSTM.
• To establish a foundation for automated predictive analysis to support disease control.

Main Methods:

• Utilized time-series analysis, multivariate data integration, and Multi-Criteria Decision Making (MCDM).
• Evaluated models based on daily confirmed cases, geographic variations, and temporal trends.
• Trained and tested models using COVID-19 data from New Zealand and its trading partners.

Main Results:

• LSTM and ARIMA models consistently outperformed Prophet.
• LSTM achieved the highest predictive accuracy, especially with 20-week datasets.
• ARIMA demonstrated superior stability and reliability for short-term forecasting.

Conclusions:

• The study identifies optimal predictive strategies for COVID-19 spread.
• Findings highlight the importance of region-specific data and training periods for model performance.
• The developed methodology enables timely, data-driven decision-making for public health authorities.