Table 4.

Comparison of the performance of BIP, GRASP, and Greedy in problem instances with 30387 candidate locations and varying number of demand points (Hollands Kroon)

Method	Demand points	Training set	Evaluation set	CPU time (hr)	Optimality gap
BIP	5000	47.34%	44.77%	0.51	<0.01%
	10000	46.86%	45.07%	2.93	<0.01%
	20000	46.30%	45.28%	28.07	<0.01%
	30000	46.05%	45.34%	41.81	0.21%
	40000	Memory error
	50000	Memory error
GRASP	5000	47.27%	44.73%	1.96
	10000	46.77%	45.05%	2.00
	20000	46.18%	45.30%	2.00
	30000	45.96%	45.31%	1.90
	40000	46.04%	45.39%	1.88
	50000	45.78%	45.41%	1.82
Greedy	5000	46.19%	43.93%	6.42e-4
	10000	46.08%	44.47%	2.21e-4
	20000	45.34%	44.68%	4.99e-4
	30000	45.29%	44.74%	7.55e-4
	40000	45.31%	44.77%	9.90e-4
	50000	45.07%	44.79%	1.25e-3