Table 5 Ablation study on composite loss components.

From: Ranking-oriented machine learning framework for probabilistic wind power forecasting with temporal reliability constraints

Model variant

MAE (MW)

RMSE (MW)

NDCG

MAP

TRSI

Proposed (full loss)

5.92

7.44

0.883

0.729

0.189

w/o ranking loss (\(\mathcal {L}_{\text {rank}}=0\))

6.65

8.01

0.826

0.665

0.204

w/o temporal reg. (\(\mathcal {L}_{\text {temp}}=0\))

6.13

7.65

0.854

0.701

0.248

w/o prediction loss (\(\mathcal {L}_{\text {pred}}=0\))

9.11

11.23

0.834

0.617

0.257

Prediction only

6.22

8.21

0.795

0.619

0.295