A combination of 'pooling' with a prediction model can reduce by 73% the number of COVID-19 (Corona-virus) tests

A method that combines 1) a neural network, for predicting whether a covid-19 is expected to be positive, based on the meta-data of the patient (i.e., their clinical symptoms if any), and 2) various test-pooling methods (a method originally suggested in the 1940's by Dorfman, but was much improved since) shows that most of the coronavirus tests can be avoided. Our analysis compares 4 known pooling methods, and shows that 2d-pooling is the best