Stock Index Prices Prediction via Temporal Pattern Attention and Long-Short-Term Memory

Table 2

Performance comparison of different methods.

Methods	Metrics	Horizon
		3	6	9	12	15	18	21	24
TPA-LSTM	test rse	0.0561	0.0931	0.1048	0.1106	0.1435	0.1309	0.1583	0.1710
	test corr	0.9792	0.9601	0.9449	0.9304	0.9166	0.9028	0.8858	0.8691
LSTNet-Skip	test rse	0.0849	0.1356	0.1663	0.2123	0.2030	0.2121	0.2091	0.2201
	test corr	0.9628	0.9338	0.9081	0.8820	0.8711	0.8536	0.8376	0.8351
LSTNet-Attn	test rse	0.0940	0.1054	0.1321	0.1516	0.1573	0.1472	0.1462	0.1786
	test corr	0.9646	0.9382	0.9234	0.9067	0.8905	0.8682	0.8578	0.8508
RNN	test rse	0.0934	0.1253	0.1379	0.1314	0.1293	0.1400	0.1545	0.1743
	test corr	0.9705	0.9522	0.9378	0.9234	0.9099	0.8877	0.8819	0.8572
CNN	test rse	0.1068	0.1961	0.3141	0.1993	0.3233	0.4418	0.3572	0.5049
	test corr	0.9530	0.9225	0.7965	0.8736	0.8740	0.7878	0.8100	0.7777
LAESN	test rse	0.0908	0.1203	0.1375	0.1740	0.1745	0.1935	0.1965	0.2474
	test corr	0.9701	0.9473	0.9332	0.9133	0.8940	0.8809	0.8570	0.8025