Modeling Markov Switching ARMA-GARCH Neural Networks Models and an Application to Forecasting Stock Returns

Table 4

Markov switching GARCH neural network models: training results.

	MSE	RMSE
Model Group 1: MS-ARMA-GARCH-neural network models
MS-ARMA-GARCH-RNN	0.035103533	0.187359368 (4th)
MS-ARMA-GARCH-RBF	0.013009577	0.114059534 (1st)
MS-ARMA-GARCH-ELMAN RNN	0.051024351	0.225885704 (5th)
MS-ARMA-GARCH-HYBRID MLP	0.034996508	0.187073536 (3rd)
MS-ARMA-GARCH-MLP	0.034665716	0.186187315 (2nd)
Model Group 2: MS-ARMA-APGARCH-neural network models
MS-ARMA-APGARCH-RNN	0.031530707	0.177568881 (4th)
MS-ARMA-APGARCH-RBF	0.056680761	0.238077218 (5th)
MS-ARMA-APGARCH-ELMAN RNN	0.027644691	0.166266929 (3rd)
MS-ARMA-APGARCH-HYBRID MLP	0.026504969	0.162803470 (1st)
MS-ARMA-APGARCH-MLP	0.026591934	0.163070336 (2nd)
Model Group 3: MS-ARMA-FIAPGARCH-neural network models
MS-ARMA-FIAPGARCH-RNN	0.029951509	0.173065042 (1st)
MS-ARMA-ARMA-FIAPGARCH-RBF	0.045969206	0.214404305 (5th)
MS-ARMA-FIAPGARCH-ELMAN RNN	0.033859273	0.184008896 (4th)
MS-ARMA-FIAPGARCH-HYBRID MLP	0.031056323	0.176228044 (2nd)
MS-ARMA-FIAPGARCH-MLP	0.031221803	0.176696926 (3rd)