From: Toward multi-label sentiment analysis: a transfer learning based approach
Model | Accuracy (%) | Hamming loss | Macro F1 | Micro F1 |
---|---|---|---|---|
Proposed models | ||||
 BERT | 87.57 | 0.011 | 0.95 | 0.96 |
 XLNet | 89.86 | 0.009 | 0.94 | 0.97 |
Baseline deep learning models | ||||
 LSTM | 76.99 | 0.021 | 0.87 | 0.92 |
 BiLSTM | 71.34 | 0.025 | 0.82 | 0.90 |
 CNN + LSTM | 76.73 | 0.021 | 0.87 | 0.92 |
Baseline machine learning models | ||||
 SGD + OR | 74.88 | 0.022 | 0.94 | 0.92 |
 LR + OR | 76.11 | 0.021 | 0.92 | 0.92 |
 SVC + OR | 80.78 | 0.017 | 0.98 | 0.94 |
 RF + OR | 73.40 | 0.023 | 0.95 | 0.92 |
 SGD + BR | 75.06 | 0.022 | 0.94 | 0.92 |
 LR + BR | 76.11 | 0.021 | 0.92 | 0.92 |
 SVC + BR | 80.78 | 0.017 | 0.98 | 0.94 |
 RF + BR | 72.07 | 0.023 | 0.95 | 0.92 |
 SGD + CC | 75.16 | 0.022 | 0.95 | 0.92 |
 LR + CC | 76.25 | 0.021 | 0.92 | 0.92 |
 SVC + CC | 80.88 | 0.017 | 0.98 | 0.94 |
 RF + CC | 72.99 | 0.024 | 0.94 | 0.91 |
 SGD + LP | 74.58 | 0.023 | 0.94 | 0.92 |
 LR + LP | 74.36 | 0.024 | 0.95 | 0.91 |
 SVC + LP | 76.08 | 0.021 | 0.98 | 0.92 |
 RF + LP | 72.64 | 0.024 | 0.97 | 0.91 |
 SGD + RakelD | 72.73 | 0.024 | 0.94 | 0.91 |
 LR + RakelD | 73.58 | 0.025 | 0.93 | 0.91 |
 SVC + RakelD | 76.42 | 0.021 | 0.98 | 0.93 |
 RF + RakelD | 72.56 | 0.024 | 0.97 | 0.91 |
 BRkNNa | 71.93 | 0.025 | 0.96 | 0.91 |
BRkNNb | 56.63 | 0.042 | 0.89 | 0.81 |
 MLARAM | 30.85 | 0.048 | 0.89 | 0.82 |
 MLkNN | 61.26 | 0.029 | 0.91 | 0.88 |