Constraint Programming for the Pandemic in Peru

Currently, the world requires techniques that match infected people and hospital beds together given various criteria such as the severity of infection, patient location, hospital capacity, etc. Deep Learning might seems to be a perfect fit for this: various configurations from a broad range of parameters that need to be reduced to a few solutions. But, this models require to be trained, hence the need for historical data on previous cases leading to a waste of time would in cleaning and consolidating a dataset and lengthy training sessions need to be performed with a variety of architectures. Nevertheless, formulating this problem as a Constraint Satisfaction Problem (CSP), the aforementioned downsides will not be present while still optimal results, and without the need for any historical data. In this paper, a CSP model is used to search for the best distribution of COVID-19 patients with a severity of patients requiring hospitalization and patients requiring ICU beds, in hospitals in a part of Lima.