Skip to main content
Financial Innovation

Table 10 MSE values of the two approaches

From: Predicting the returns of the US real estate investment trust market: evidence from the group method of data handling neural network

Training/testing

GARCH Model

GMDH Model

Linear kernel

Linear covariance kernel

Quadratic kernel

30/30

4.9346

3.69

57.22

–

100/30

3.7935

0.72

1.05

–

200/30

2.2102

0.70

0.69

–

300/30

1.8723

1.45

1.08

4.35

600/30

1.6315

0.87

–

0.62

300/60

1.9317

1.93

0.82

–

  1. Note: The training and testing numbers of samples for the GARCH model indicate the number of samples used for calculating the coefficients and then the number of returns predicted by the model. "–" indicates that these cases are not accepted in this study because their predicted results are greatly impacted by outliers
Back to article page