Forecasting Thailand's mobility trends using Feature Engineered XGBoost for pandemic crisis movement management

Area of Science:

• Epidemiology
• Data Science
• Public Health

Background:

• The COVID-19 pandemic caused significant global mobility changes, impacting virus transmission.
• Google Community Mobility Reports offer anonymized location data to track movement trends.
• Mobility insights are vital for timely public health interventions and data-driven decision-making.

Purpose of the Study:

• To forecast human mobility trends in Thailand during the COVID-19 pandemic.
• To evaluate the performance of different forecasting models using Google's mobility data.
• To examine the relationship between mobility patterns and COVID-19 cases under lockdown conditions.

Main Methods:

• Applied three forecasting models: Facebook Prophet, ARIMA, and Feature Engineered XGBoost.
• Utilized Google Community Mobility Reports data for Thailand.
• Employed the Granger Causality Test to analyze mobility and COVID-19 case correlations during lockdown phases.

Main Results:

• Feature Engineered XGBoost achieved the highest accuracy in forecasting mobility trends across all six location categories.
• The Granger Causality Test revealed relationships between mobility patterns and COVID-19 case numbers during different lockdown phases.
• The study confirmed the utility of machine learning for predicting mobility during public health restrictions.

Conclusions:

• Machine learning models effectively forecast human mobility trends during public health restrictions.
• Understanding mobility patterns is critical for disease transmission control.
• Insights can inform targeted mobility management policies and public health responses for future outbreaks.