Table 3 Resilience of the xFiTRNN model on both financial phrasebank and IMBSEntFiN datasets.
Dataset | Model | A | Macro Avg P | Macro Avg R | Macro Avg F1 | Macro avg AUC | Train time (s) | Infer time (s) |
|---|---|---|---|---|---|---|---|---|
Financial Phrasebank | 1 Hidden Layer NN | 89.30 | 86.50 | 92.00 | 89.00 | 93.04 | 1494.30 | 90.25 |
2 Hidden Layers NN | 90.40 | 87.70 | 92.90 | 90.16 | 94.16 | 2000.75 | 119.15 | |
3 Hidden Layers NN | 91.80 | 88.90 | 94.30 | 91.24 | 95.28 | 2464.20 | 101.30 | |
BiGRU + 3 Hidden Layers NN | 93.40 | 90.60 | 95.60 | 92.74 | 96.60 | 3000.85 | 110.45 | |
BiGRU + CNN | 94.00 | 91.10 | 96.20 | 93.32 | 96.99 | 3500.35 | 130.20 | |
Proposed model (xFiTRNN) | 97.61 | 97.33 | 97.38 | 97.48 | 98.58 | 3857.21 | 165.41 | |
IMBSEntFiN | 1 Hidden Layer NN | 87.70 | 85.50 | 89.60 | 87.58 | 91.04 | 870.50 | 58.30 |
2 Hidden Layers NN | 88.80 | 86.50 | 90.90 | 88.56 | 92.16 | 1166.30 | 53.15 | |
3 Hidden Layers NN | 89.60 | 87.50 | 91.70 | 89.64 | 93.28 | 1400.75 | 46.80 | |
BiGRU + 3 Hidden Layers NN | 90.80 | 88.60 | 93.20 | 90.94 | 94.20 | 2200.70 | 140.48 | |
BiGRU + CNN | 91.20 | 89.10 | 93.60 | 91.32 | 94.79 | 2502.40 | 120.10 | |
Proposed model (xFiTRNN) | 94.11 | 93.83 | 93.88 | 93.98 | 95.08 | 2200.00 | 94.25 |
