Skip to main content

Table 8 Model configuration for the best model in the entire period for classical analysis

From: A hybrid econometrics and machine learning based modeling of realized volatility of natural gas

Ranking

Model

Financial assets

NoL

NpL

WS

MSE

RMSE

MAE

1

LSTM

BRENT-SPX

1

150

1000

2,38E−01

4,88E-01

8,85E−04

2

LSTM

BRENT-XAU

1

150

1000

2,38E−01

4,88E−01

8,64E−04

3

LSTM

BRENT-XAU-SPX

1

150

700

2,38E−01

4,88E−01

8,27E−04

4

LSTM

BRENT-EURO-XAU

2

150

700

2,38E−01

4,88E−01

9,31E−04

5

LSTM

BRENT

1

150

700

2,38E−01

4,88E−01

8,18E−04

 

HAR

SPX

  

700

2,43E−01

4,93E−01

8,35E−04

  1. In bold is the best model by loss function criteria.
  2. NoL Number of layers, NpL Neurons per layer, WS Window size.