Deep Learning for Price Movement Prediction Using Convolutional Neural Network and Long Short-Term Memory

Table 4

The macroaverage F-measure of CNN3D-DR + LSTM with different parameters on validation set in S&P 500 when .

Number of hidden neurons	Dropout rate	Time steps in LSTM
		n = 6	n = 7	n = 8	n = 9	n = 10	n = 11	n = 12	n = 13	n = 14	n = 15
N = 50	0.1	0.6289	0.6107	0.6213	0.5865	0.5583	0.5912	0.5708	0.5546	0.5368	0.5767
	0.2	0.6057	0.6055	0.5928	0.5759	0.5565	0.6254	0.5720	0.5888	0.5787	0.5492
	0.3	0.6177	0.6070	0.5801	0.6235	0.5943	0.6059	0.5640	0.5622	0.5445	0.5445
	0.4	0.6147	0.6207	0.5822	0.6070	0.5604	0.6131	0.5593	0.5599	0.5559	0.5582
N = 100	0.1	0.6095	0.6033	0.6156	0.5980	0.5883	0.6223	0.5739	0.5572	0.5467	0.5380
	0.2	0.5879	0.6069	0.5792	0.5874	0.5988	0.6442	0.5753	0.5821	0.5783	0.5505
	0.3		0.6190	0.5980	0.5950	0.5540	0.6217	0.5723	0.5993	0.5545	0.5555
	0.4	0.5998	0.6107	0.5996	0.5950	0.5681	0.6028	0.5681	0.5991	0.5405	0.5593
N = 150	0.1	0.6095	0.6295	0.5950	0.5794	0.5644	0.5884	0.5811	0.5555	0.5559	0.5613
	0.2	0.6170	0.6010	0.5876	0.6021	0.5876	0.6258	0.5940	0.5458	0.5915	0.5622
	0.3	0.6057	0.6093	0.5876	0.6137	0.5816	0.6187	0.5786	0.5678	0.5368	0.5626
	0.4	0.5863	0.6123	0.5749	0.6026	0.5597	0.6203	0.5487	0.5686	0.5661	0.5603
N = 200	0.1	0.6043	0.6198	0.6040	0.5809	0.5712	0.6284	0.5375	0.5661	0.5456	0.5245
	0.2	0.5981	0.6174	0.5869	0.6033	0.5621	0.5919	0.5708	0.5858	0.5467	0.5670
	0.3	0.6035	0.6093	0.6047	0.5988	0.5943	0.6150	0.5895	0.5970	0.5268	0.5513
	0.4	0.6078	0.6204	0.5869	0.5853	0.5742	0.6269	0.5357	0.5670	0.5473	0.5783