Construction heuristic of initial population

[martinwholtmon]

Initialization takes forever, so must introduce a construction heuristic for the initial population.

Thinking of using KNN with the data points to generate clusters of patients.

• [ ] Assign n patients to clusters dependent on
  • Calculate cluster demand
  • Check for patient overlap (number of concurrent overlaps).
    • Optimistic approach
    • Check the time windows and try to arrange the patients accordingly. Do we need the same amount of nurses as overlaps, or could we fit in the patients such that less nurses can attend to them?
• [ ] Calculate cluster's start- and endtime so nurses of other clusters can be assigned if necessary.

Constraints

• Check for constraints within a cluster
• If cluster is not ok, halt and wait for other clusters to finish. Then try to assign an available nurse from another cluster
• If all clusters are ok = new populations

Tips:

• one thread per cluster