The total net benefit of 4,128 sequential treatment policies was examined through the underlying evaluation model (see Figure 6.). The outer loop generated a sequential treatment policy, Policy, and sent the policy to the underlying evaluation model ‘EvModel’ consecutively. The inner loop repeated 100 times to produce the total net benefits, TNB, of each policy. The optimal solution was determined based on the mean of the 100 sampled total net benefits. Due to the computational burden to run 4,128 policies for 100 times, enumeration was implemented through Iceberg, which is the Linux based High Performance Computing Cluster at the University of Sheffield. 100 samples, which were independent of each other, were run in parallel using two of Intel X5650 6-core processors (12 cores in total).
matlabpool open12 % Open 12 workers.
FOR Policy = 1:4128 % Policy generation.
% Repeat PSAs for 100 times in parallel using 12 workers.
