An Improved Particle Swarm Optimization for Selective Single Machine Scheduling with Sequence Dependent Setup Costs and Downstream Demands
Table 5
Comparison results for different kinds of PSO algorithms.
CPLEX
PSOVNS
PSOadaptiveVNS
Gap (%)
CPU (s)
Gap (%)
CPU (s)
Gap (%)
Gains%
CPU (s)
3
50
0.00
32.54
0.00
17.01
0.00
0.00
12.74
80
0.00
48.75
0.06
87.35
0.06
0.00
71.35
100
1.14
1637.63
0.46
150.00
0.46
0.00
150.00
120
0.95
1615.27
0.45
180.00
0.41
8.89
180.00
150
2.56
1608.52
1.42
225.00
1.34
5.63
225.00
200
5.39
1605.28
2.86
300.00
5.94
300.00
5
50
0.00
74.07
0.00
27.76
0.00
0.00
22.06
80
0.09
581.50
0.17
200.00
0.17
0.00
200.00
100
1.68
1617.86
0.68
250.00
0.62
8.82
250.00
120
2.36
1612.21
1.24
300.00
1.15
7.26
300.00
150
2.81
1607.12
1.79
375.00
7.26
375.00
200
6.86
1603.56
3.66
500.00
8.74
500.00
7
50
0.00
71.64
0.02
26.54
0.02
0.00
24.86
80
1.04
1606.09
0.94
280.00
0.86
8.51
280.00
100
2.77
1624.07
1.47
350.00
8.84
350.00
120
2.51
1602.53
1.39
420.00
1.31
5.76
420.00
150
4.36
1609.16
2.36
525.00
16.53
525.00
200
8.42
1615.82
4.58
700.00
8.73
700.00
Denotes that the performance difference between PSOadaptiveVNS and PSOVNS is significantly for a certain problem size based on the Pairwise-Samples Test with a confidence level of 95%.
