Table 4

Optimized hyperparameters of ANN of combination models.

Hidden layer node number (CME/CIR) Learning rate (CME/CIR) Learning cycle (CME/CIR)
1-hour prediction 4/4 0.01/0.01 1500/1500
2-hour prediction 24/55 0.01/0.01 1400/800
3-hour prediction 5/58 0.01/0.001 1300/1500
4-hour prediction 5/45 0.01/0.001 1300/1500
5-hour prediction 6/43 0.01/0.001 1500/1400
6-hour prediction 5/17 0.01/0.001 1300/1500
7-hour prediction 2/24 0.001/0.001 1400/200
8-hour prediction 2/25 0.001/0.002 1400/1200
9-hour prediction 2/25 0.001/0.003 1400/1200
10-hour prediction 2/25 0.001/0.004 1400/1200
11-hour prediction 2/1 0.001/0.0001 1400/1100
12-hour prediction 2/1 0.001/0.0001 1400/1100
13-hour prediction 32/1 0.0001/0.0001 500/1100
14-hour prediction 37/1 0.0001/0.0001 500/1100
15-hour prediction 37/1 0.0001/0.0001 500/1100
16-hour prediction 37/58 0.0001/0.001 500/1500
17-hour prediction 37/60 0.0001/0.001 500/1500
18-hour prediction 37/60 0.0001/0.001 500/1500
19-hour prediction 37/1 0.0001/0.0001 500/1100
20-hour prediction 37/1 0.0001/0.0001 500/1100
21-hour prediction 37/1 0.0001/0.0001 500/1100
22-hour prediction 37/1 0.0001/0.0001 500/1100
23-hour prediction 37/1 0.0001/0.0001 500/1100
24-hour prediction 37/1 0.0001/0.0001 500/1100

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.