Meta-Learning Enhanced Trade Forecasting: A Neural Framework Leveraging Efficient Multicommodity STL Decomposition

Table 4

Performance comparison of various models for exported commodity value series.

Model	MAE	RMSE	MAPE
LastValuePredictor	3.50	7.36	0.1854
ARIMA	2.64	5.29	0.1811
VAR	7.15	8.13	0.3802
Bagging	3.63	5.14	0.1983
LSTM	3.70	7.18	0.2064
GRU	3.82	8.15	0.2344
DeepAR	3.00	6.60	0.2210
DeepVAR	2.65	5.96	0.2185
N-Beats	2.92	6.08	0.1763
N-Hits	3.04	6.77	0.1804
TFT	2.76	4.89	0.1781
TFSTL
Meta-TFSTL (fine-tuning)

Bold: best; underline: second best; italics: best in baseline.