CNN-LSTM Model Optimized by Bayesian Optimization for Predicting Single-Well Production in Water Flooding Reservoir

Table 5

Comparison of prediction accuracy of different models under different feature selection strategies (84-3).

Models (84-3)		MAE			MAPE/%			Accurate/%
		MAE	Max	Min	MAPE	Max	Min
A	CNN-LSTM	17.60	123.1	0.24	6.55	34.2	0.04	93.30	0.99
	CNN	22.66	125.31	0.21	7.19	38.1	0.03	88.37	0.99
	LSTM	23.55	120.13	0.19	7.32	35.1	0.03	86.11	0.99
	GRU	22.45	121.45	0.23	7.31	37.21	0.04	85.34	0.99
B	CNN-LSTM	17.58	120.03	0.23	6.54	35.4	0.03	93.23	0.99
	CNN	22.65	125.47	0.22	7.73	37.9	0.03	88.36	0.99
	LSTM	20.33	120.15	0.2	7.02	34.1	0.03	90.01	0.99
	GRU	21.45	121.63	0.24	7.15	34.2	0.04	90.24	0.99

A: use all features; B: use selected features.