Multivariate and Online Prediction of Closing Price Using Kernel Adaptive Filtering

Table 4

Best embedding dimension for all time windows according to evaluation metrics MSE, MAE, and DS (2021).

Time window	MSE	Best embedding dimensions	Algorithms
1 day	0.034 2	2	KRLS
60 minutes	0.008 1	2	KRLS
30 minutes	0.004 7	2	KRLS
25 minutes	0.004 2	2	KRLS
20 minutes	0.003 3	2	KRLS
15 minutes	0.002 8	2	KRLS
10 minutes	0.002 0	2	KRLS
5 minutes	0.001 1	2	KRLS
1 minute	0.000 3	2	KRLS
Time window	MAE	Best embedding dimensions	Algorithms
1 day	1.690 1	2	KRLS
60 minutes	0.548 0	2	KRLS
30 minutes	0.396 1	2	KRLS
25 minutes	0.367 1	2	KRLS
20 minutes	0.323 8	2	KRLS
15 minutes	0.280 3	2	KRLS
10 minutes	0.225 9	2	KRLS
5 minutes	0.160 1	2	KRLS
1 minute	0.072 9	2	KRLS
Time window	DS	Best embedding dimensions	Algorithms
1 day	0.487 0	4	NORMA
60 minutes	0.493 0	4	KNLMS
30 minutes	0.484 9	4	KRLS
25 minutes	0.487 8	6	LKAPA
20 minutes	0.489 1	7	LKAPA
15 minutes	0.488 1	2	PROB-LMS
10 minutes	0.489 1	2	PROB-LMS
5 minutes	0.484 6	2	PROB-LMS
1 minute	0.475 1	2	KRLS