A Hybrid LSTM-Based Ensemble Learning Approach for China Coastal Bulk Coal Freight Index Prediction

Table 5

Predictive performance of hybrid LSTM-EL and benchmark models for CBCFI prediction.

Methods	MAPE (%)	RMSE	MASE	Dsta
Daily CBCFI forecasting; Training period: 2012/1/4∼2017/4/10; Testing period: 2019/1/7∼2020/10/13
LSTM	9.15	0.2419	1.1532	0.5703
GBRT	7.84	0.2342	0.8915	0.5145
LSTM-GBRT	6.00	0.1369	0.9132	0.8893
RF	8.62	0.2034	0.8370	0.5042
LSTM-RF	6.50	0.1531	0.9560	0.8539
Weekly CBCFI forecasting; Training period: 2012/1/9∼2017/3/24; Testing period: 2019/1/21∼2020/9/28
LSTM	9.47	0.1931	1.6532	0.4303
GBRT	8.99	0.1815	0.6173	0.5324
LSTM-GBRT	4.08	0.1232	0.9835	0.8034
RF	7.32	0.1673	0.5525	0.4853
LSTM-RF	4.43	0.1476	0.8941	0.8113
Monthly CBCFI forecasting; Training period: 2012/1/4∼2017/3/4; Testing period: 2019/4/1∼2020/10/13
LSTM	11.53	0.2571	1.0126	0.5863
GBRT	10.89	0.2212	0.7940	0.5932
LSTM-GBRT	5.04	0.1099	0.9203	0.9024
RF	11.01	0.2236	0.7864	0.6072
LSTM-RF	4.25	0.0943	0.8825	0.9043