Forecasting directional movement of Forex data using LSTM with technical and macroeconomic indicators

Financial Innovation

Table 18 Hybrid model (on extended dataset): three-days-ahead predictions

Iterations		Hybrid model-modification based on ME_LSTM		Hybrid model-modification based on TI_LSTM
ME_LSTM	TI_LSTM	Profit_accuracy (%)	# of transactions	Profit_accuracy (%)	# of transactions
50	50	60.71	56/152	57.14	56/152
50	100	71.43	42/152	66.67	42/152
50	150	51.79	56/152	58.93	56/152
50	200	100.00	2/152	100.00	2/152
100	50	63.86	83/152	53.01	83/152
100	100	73.02	63/152	61.90	63/152
100	150	58.11	74/152	58.11	74/152
100	200	86.67	55/152	88.68	55/152
150	50	82.00	50/152	74.00	50/152
150	100	65.00	40/152	67.50	40/152
150	150	62.50	48/152	64.58	48/152
150	200	81.25	37/152	82.35	37/152
200	50	88.89	10/152	88.89	10/152
200	100	83.33	6/152	83.33	6/152
200	150	57.14	8/152	62.50	8/152
200	200	62.50	8/152	57.14	8/152
Average		71.76	39.88/152	70.30	39.88/152