Multivariate and Online Prediction of Closing Price Using Kernel Adaptive Filtering

Table 3

Best embedding dimension for all time windows according to evaluation metrics MSE, MAE, and DS (2020).

Time window	MSE	Best embedding dimensions	Algorithms
1 day	0.014 3	2	KRLS
60 minutes	0.003 4	2	KRLS
30 minutes	0.002 0	2	KRLS
25 minutes	0.001 8	2	KRLS
20 minutes	0.001 5	2	KRLS
15 minutes	0.001 2	2	KRLS
10 minutes	0.000 8	2	KRLS
5 minutes	0.000 4	2	KRLS
1 minute	0.000 10	2	KRLS
Time window	MAE	Best embedding dimensions	Algorithms
1 day	2.132 4	2	KRLS
60 minutes	0.660 2	2	KRLS
30 minutes	0.466 6	2	KRLS
25 minutes	0.431 7	2	KRLS
20 minutes	0.381 2	2	KRLS
15 minutes	0.329 5	2	KRLS
10 minutes	0.266 7	2	KRLS
5 minutes	0.188 6	2	KRLS
1 minute	0.085 2	2	KRLS
Time window	DS	Best embedding dimensions	Algorithms
1 day	0.501 3	5	LKAPA
60 minutes	0.491 9	5	KRLS
30 minutes	0.491 3	5	KRLS
25 minutes	0.492 5	3	KRLS
20 minutes	0.488 1	2	PROB-LMS
15 minutes	0.489 3	7	QKLMS
10 minutes	0.490 2	2	KRLS
5 minutes	0.491 0	2	PROB-LMS
1 minute	0.471 5	2	KRLS