Novel FTLRNN with Gamma Memory for Short-Term and Long-Term Predictions of Chaotic Time Series

Table 21

For K = 24, training and testing samples variation for FTLRNN on testing data set for monthly sunspot time series and for laser time series (K = 50).

Time series		Sunspot time series 𝐾=24			Laser time series 𝐾=50
Training exemplars	Testing Exemplars	MSE	NMSE	𝑟	MSE	NMSE	𝑟
10%	75%	0.03246	0.37734	0.49224	0.04073	1.18215	0.13264
20%	65%	0.03657	0.58297	0.43616	0.04310	1.23643	-0.00980
30%	55%	0.02280	0.52882	0.69373	0.03018	0.98884	0.27393
40%	45%	0.01839	0.83622	0.77659	0.04425	1.87215	-0.22273
50%	35%	0.04066	0.41935	0.53973	0.02534	0.94845	0.26601
60%	25%	0.03019	0.54899	0.71497	0.03466	1.02971	0.41330
70%	15%	0.02834	0.95150	0.75528	0.03447	0.89215	0.44330
80%	05%	0.02339	0.57734	0.8012	0.08839	0.96548	0.28132