Predicting Travel Patterns to Healthcare Facilities for Delaware Communities During Flooding

• Jack Kingham, Computer Science, University of Delaware

• Shangjia Dong, Civil and Environmental Engineering, University of Delaware

Ensuring access to critical facilities such as dialysis or childcare facilities during disasters is the premise of building a resilient community. Thus, understanding the patterns of how individuals visit critical facilities will enable more effective city planning and optimal disaster relief strategies. However, acquiring each person’s travel trajectory can be very expensive and raises privacy concerns.  By harnessing advanced traffic methods and human mobility big data, this paper employs the gravity model for trip prediction to anticipate the travel patterns of the Delaware population to dialysis and childcare facilities at a census block group level. Population attributes such as age distribution, sociodemographics, and population centrality are used to better inform the model’s predictions. To validate the predicted travel patterns, they are compared to real travel patterns extracted from the SafeGraph Patterns dataset. Particle swarm optimization is then used to optimize the gravity model and find ideal parameters that produce travel patterns with the greatest correlation to the SafeGraph data. Finally, using synthetic population and flood depth data, the predicted travel patterns are redistributed at a household level to identify individuals who visit dialysis and childcare facilities, and their travel patterns to the facilities they visit. Integration of hazard risks and facility visit prediction will enable informed service provision and targeted interventions.