Research Article
Hybrid Deep Neural Network Scheduler for Job-Shop Problem Based on Convolution Two-Dimensional Transformation
Table 8
Comparison of HDNNS with other methods on ZLP datasets.
| | | Optimal | HDNNS (our) | DQN | DEEPRM | HNN | ANN (1D) | ANN (all) | STPT | SPT |
| ZLP (8 8)
pU (10, 20) | Average MAKESPAN | 156.82 | 174.25 | 177.80 | 178.51 | 190.52 | 195.52 | 182.52 | 181.58 | 201.11 | | Scheduling score | — | 90.0 | 88.2 | 87.9 | 82.3 | 80.2 | 85.9 | 86.4 | 78.0 | | Training time | — | 5396.2 | 12987.3 | 12568.6 | — | 427.6 | 9482.5 | — | — | | Scheduling time | — | 292.7 | 298.7 | 294.3 | 352.6 | 407.5 | 367.4 | 230.1 | 200.1 |
| ZLP (8 8)
pU (10, 40) | Average MAKESPAN | 269.93 | 305.69 | 306.39 | 309.05 | 331.85 | 338.21 | 313.54 | 324.19 | 347.45 | | Scheduling score | — | 88.3 | 88.1 | 87.3 | 81.3 | 79.8 | 86.1 | 83.3 | 77.7 | | Training time | — | 5124.5 | 13974.2 | 13213.8 | — | 425.8 | 9426.6 | — | — | | Scheduling time | — | 291.5 | 295.3 | 293.5 | 353.6 | 408.4 | 392.1 | 210.2 | 210.2 |
| ZLP (8 8)
pU (10, 60) | Average MAKESPAN | 386.43 | 428.41 | 436.15 | 434.14 | 458.84 | 486.07 | 449.62 | 464.49 | 488.51 | | Scheduling score | — | 90.2 | 88.6 | 89.0 | 84.2 | 79.5 | 85.9 | 83.2 | 79.1 | | Training time | — | 5135.8 | 13488.82 | 12846.5 | — | 424.5 | 9814.2 | — | — | | Scheduling time | — | 264.6 | 280.3 | 293.5 | 371.6 | 401.1 | 391.4 | 250.3 | 180.5 |
| ZLP (8 8)
pU (10, 80) | Average MAKESPAN | 502.89 | 562.52 | 570.81 | 561.26 | 623.69 | 591.64 | 590.15 | 593.12 | 632.95 | | Scheduling score | — | 89.4 | 88.1 | 89.6 | 80.6 | 85.0 | 85.2 | 84.8 | 79.5 | | Training time | — | 5217.7 | 13425.7 | 12681.6 | — | 425.4 | 9823.6 | — | — | | Scheduling time | — | 271.2 | 273.4 | 286.5 | 401.4 | 406.7 | 408.5 | 220.6 | 200.1 |
| ZLP (13 13)
pU (10, 20) | Average MAKESPAN | 290.58 | 332.85 | 332.47 | 331.63 | 357.67 | 345.92 | 338.69 | 380.97 | 402.98 | | Scheduling score | — | 87.3 | 87.4 | 87.6 | 81.2 | 84.0 | 85.8 | 76.3 | 72.1 | | Training time | — | 19162.9 | 61588.4 | 58584.7 | — | 424.0 | 37568.1 | — | — | | Scheduling time | — | 2384.5 | 2634.9 | 2871.5 | 5371.6 | 407.7 | 2589.4 | 2151.1 | 1840.1 |
| ZLP (13 13)
pU (10, 40) | Average MAKESPAN | 500.77 | 551.50 | 582.96 | 570.02 | 634.55 | 589.14 | 575.16 | 669.81 | 700.25 | | Scheduling score | — | 90.8 | 85.9 | 87.9 | 78.9 | 85.0 | 87.1 | 74.8 | 71.5 | | Training time | — | 20540.5 | 60974.3 | 58071.0 | — | 420.5 | 37481.6 | — | — | | Scheduling time | — | 2372.5 | 2628.4 | 2899.8 | 5375.4 | 407.0 | 2648.6 | 2284.4 | 1863.8 |
| ZLP (13 13)
pU (10, 60) | Average MAKESPAN | 720.57 | 794.45 | 819.76 | 814.20 | 905.10 | 885.22 | 869.26 | 868.37 | 913.87 | | Scheduling score | — | 90.7 | 87.9 | 88.5 | 79.6 | 81.4 | 82.9 | 83.0 | 78.8 | | Training time | — | 19346.4 | 64789.2 | 60325.7 | — | 425.1 | 39426.4 | — | — | | Scheduling time | — | 2384.8 | 2677.2 | 2987.7 | 5471.3 | 406.7 | 2468.7 | 2145.6 | 2056.1 |
| ZLP (13 13)
pU (10, 80) | Average MAKESPAN | 1026.57 | 1123.16 | 1210.57 | 1203.90 | 1311.87 | 1262.69 | 1255.47 | 1219.00 | 1299.63 | | Scheduling score | — | 91.4 | 84.8 | 85.3 | 78.3 | 81.3 | 81.8 | 84.2 | 79.0 | | Training time | — | 19741.3 | 60148.5 | 58247.7 | — | 425.8 | 40259.6 | — | — | | Scheduling time | — | 2346.9 | 2615.5 | 2884.1 | 5577.7 | 407.7 | 2945.4 | 2181.5 | 2054.1 |
| | AVE | Average MAKESPAN | 481.82 | 534.10 | 554.61 | 550.34 | 601.76 | 586.80 | 571.80 | 587.69 | 623.34 | | Scheduling score | — | 89.8 | 87.3 | 87.9 | 80.8 | 82.0 | 85.1 | 82.0 | 77.0 | | Training time | — | 12458.2 | 37672.1 | 35817.5 | — | 424.8 | 24160.3 | — | — | | Scheduling time | — | 1556.1 | 1579.9 | 1601.4 | 2909.4 | 406.6 | 1526.4 | 1209.2 | 1075.6 |
| | Average MAKESPAN rank | 1 | 3 | 2 | 7 | 5 | 4 | 6 | 8 | | Training time rank | 2 | 4 | 3 | — | 1 | 5 | — | — | | Scheduling time RANK | 5 | 6 | 7 | 8 | 1 | 4 | 3 | 2 |
|
|
