Table 5.
Comparison of Mean PMAE between MPark Methods and the No-Change Model
| Year 1 | Year 2 | Year 3 | Year 4 | Year 5 | |
|---|---|---|---|---|---|
| No-change | 2.33 ± 0.027 | 2.52 ± 0.071 | 2.79 ± 0.222 | 2.34 ± 0.292 | 3.06 ± 0.650 |
| MPark | 2.37 ± 0.036 | 2.55 ± 0.112 | 2.82 ± 0.252 | 2.42 ± 0.358 | 3.17 ± 0.844 |
| MPark (oversampling) | 2.68 ± 0.032 | 2.85 ± 0.102 | 3.10 ± 0.229 | 2.85 ± 0.322 | 3.27 ± 0.802 |
| MPark (undersampling) | 3.98 ± 0.142 | 4.38 ± 0.333 | 4.29 ± 0.941 | 6.24 ± 1.590 | 5.97 ± 2.280 |
MPark = recurrent neural network method from Park et al; PMAE = pointwise mean absolute error.