CRII: AF: Machine Learning Models and Optimization Algorithms for Management of Volatile Public Health Emergencies and Disasters

Efficient management of emergency response logistics is vital for impact mitigation, outbreak containment, and targeted intervention immediately following a sudden epidemic outbreak or a fast-changing disaster. Pre-planning activities include optimal positioning of Emergency Operation Centers and Points of Dispensing to ensure prompt response. This project seeks to investigate and develop machine learning models and optimization algorithms for determining the minimal number of depots, center locations, and the volume of interventions or life-saving equipment at each center. It optimizes emergency response logistics through the development of models that satisfy the immediate needs of the affected rural or semi-urban population in a quick-onset and fast-changing bio-medical disaster or public health emergency. In addition to these models, a novel algorithm will be developed that will will offer improvement in understanding time complexity, while ensuring stable and efficient output. The algorithm uses the capacity (number of interventions required to service the regional population) and duration (measure of distance and road infrastructure from the depot to the affected regions). Specifically, the proposal will develop a real-time algorithm and decision support system for optimizing disaster response logistics. Application of this real-time analyses is envisaged to positively impact communities, emergency response agencies, officials and planners (particularly in low- and middle-income countries) by reducing the cost of response operations and the number of casualties whenever there is a sudden disease outbreak. This project will advance theoretical knowledge of combinatorial optimization and its application to finding solutions for real-life problems. It increases the scalability of methods for solving computationally NP-Hard problems. The project aids the discovery of relevant clustering techniques while fostering further innovations for the future STEM projects. It uses data from disparate sources: population, environmental, geospatial, and resource availability. Specifically, the project seeks to accomplish the following objectives: research existing machine learning models for clustering regions and their time complexities, with a view to developing an enhanced K-means ++ clustering model with improved time complexity; develop a novel Dual-Determinant Clustering Algorithm that involves using the regional capacity and distance (which is a factor of the Haversine distance and road infrastructure from the depot to the endpoints/regions. The new algorithm will be used to optimize post-disaster response operations and logistics; develop real-time algorithms and decision support system for optimizing routing of response logistics during disease outbreaks. This will include models for dynamic integration of locations with fuzzy demands and uncertainties. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.