Skip to main content
. 2024 Feb 16;10:e1859. doi: 10.7717/peerj-cs.1859

Table 10. Comparison of fine-tuned BERT with benchmarks for multi-class classification (precision and recall).

Disasters Model Precision Recall
Infra Human Non Macro Infra Human Non Macro
Iraq–Iran Earthquake Rudra et al. (2018) 0 74.53 56.99 43.84 0 66.92 77.73 48.22
Word2vec+RF 57.14 76.32 80.00 71.15 38.10 93.55 47.06 59.57
TF-IDF+SVM 99.0 88.89 86.25 91.71 16.67 66.67 99.0 60.78
TF-IDF+LSTM 79.12 78.69 87.64 81.81 40.89 84.79 92.55 72.74
TF-IDF+Bi-LSTM 82.38 81.21 88.24 83.94 43.19 85.58 92.60 73.79
TF-IDF+CNN 83.15 82.76 89.74 85.21 45.10 86.83 92.68 74.86
Madichetty & Sridevi (2021) 88.33 79.88 74.50 80.90 66.67 80.66 77.27 74.87
Proposed 66.67 95.00 94.55 85.40 50.00 97.44 94.55 80.66
Sri Lanka Floods Rudra et al. (2018) 42.08 63.33 64.94 56.78 44.50 70.50 37.50 50.83
Word2vec+RF 57.14 76.32 80.00 71.15 38.10 93.55 47.06 59.57
TF-IDF+RF 50.00 87.50 81.93 73.14 25.00 56.00 97.14 59.38
TF-IDF+LSTM 58.45 87.90 82.69 76.34 41.49 84.10 88.34 71.31
TF-IDF+Bi-LSTM 59.78 88.24 83.87 77.29 43.58 84.58 89.46 72.54
TF-IDF+CNN 60.42 88.69 84.79 77.96 45.69 85.69 90.67 74.01
Madichetty & Sridevi (2021) 85.33 84.83 78.02 82.73 70.50 70.50 93.00 78.0
Proposed 61.11 89.47 99.20 83.26 100.00 94.44 93.23 95.89
Mexico Earthquake Rudra et al. (2018) 50.59 69.00 50.32 39.77 0 59.61 85.98 48.53
TF-IDF+SVM 87.50 90.91 91.63 90.01 38.89 68.97 99.49 69.12
Word2vec+RF 57.14 76.32 80.00 71.15 38.10 93.55 47.06 59.57
TF-IDF+LSTM 59.38 77.71 81.49 72.86 39.56 93.78 47.43 60.25
TF-IDF+Bi-LSTM 59.85 78.32 82.38 73.51 40.32 93.97 49.39 61.22
TF-IDF+CNN 60.12 79.78 83.58 74.49 41.58 94.12 51.78 62.49
Madichetty & Sridevi (2021) 66.02 73.05 67.71 68.93 58.27 70.16 74.25 67.56
Proposed 53.33 86.67 94.39 78.13 61.54 59.09 96.65 72.43
California Wildfires Rudra et al. (2018) 33.15 54.16 72.78 53.36 51.00 50.32 43.76 48.36
TF-IDF+SVM 71.43 100.0 84.59 85.34 16.13 57.89 99.56 57.86
Word2vec+RF 57.14 76.32 80.00 71.15 38.10 93.55 47.06 59.57
TF-IDF+LSTM 72.59 99.10 85.19 85.62 51.29 70.45 82.87 68.20
TF-IDF+Bi-LSTM 73.39 99.31 86.49 86.39 51.78 73.59 83.29 69.55
TF-IDF+CNN 74.19 99.43 87.69 87.10 52.10 75.98 85.48 71.18
Madichetty & Sridevi (2021) 59.98 72.14 57.67 63.26 51.69 66.87 66.95 61.84
Proposed 54.55 75.41 90.50 73.49 52.94 90.20 86.60 76.58
Hurricane Harvey Rudra et al. (2018) 42.42 46.22 49.98 46.21 73.23 7.92 40.06 40.40
TF-IDF+SVM 65.52 85.71 88.76 80.00 28.36 13.64 98.53 46.84
Word2vec+RF 57.14 76.32 80.00 71.15 38.10 93.55 47.06 59.57
TF-IDF+LSTM 66.59 76.57 81.23 74.79 39.49 93.71 48.39 60.53
TF-IDF+Bi-LSTM 66.74 76.75 81.46 74.98 40.36 93.85 50.25 61.48
TF-IDF+CNN 67.21 77.04 81.76 75.33 41.59 94.10 51.59 62.42
Madichetty & Sridevi (2021) 73.34 64.31 67.17 68.27 64.97 55.25 79.55 66.59
Proposed 53.98 63.49 94.32 70.60 71.76 71.43 89.25 77.48
Hurricane Maria Rudra et al. (2018) 45.15 0 53.63 32.97 76.90 0 40.34 39.08
Word2vec+RF 60.00 76.62 72.73 69.78 40.91 93.65 44.44 59.67
TF-IDF+SVM 77.78 100.0 91.83 89.87 14.00 12.50 99.72 42.07
TF-IDF+LSTM 78.10 98.21 91.95 89.42 18.45 20.38 90.10 42.97
TF-IDF+Bi-LSTM 78.48 98.43 92.13 89.68 20.27 22.87 90.56 44.56
TF-IDF+CNN 78.69 98.65 92.37 89.90 32.78 25.76 92.18 50.24
Madichetty & Sridevi (2021) 77.16 71.17 74.73 74.35 75.52 48.63 85.52 69.89
Proposed 47.37 42.86 97.69 62.64 54.55 69.23 95.99 73.26
Hurricane Irma Rudra et al. (2018) 50.17 0 57.10 35.76 73.95 0 52.47 42.14
TF-IDF+SVM 61.11 57.14 88.31 68.86 14.86 11.43 99.13 41.81
Word2vec+RF 56.25 76.32 61.54 64.70 40.91 89.23 44.44 58.19
TF-IDF+LSTM 61.49 77.38 88.53 75.80 42.41 88.49 44.78 58.56
TF-IDF+Bi-LSTM 62.21 77.79 89.11 76.37 43.89 89.34 45.29 59.50
TF-IDF+CNN 62.54 78.10 89.34 76.66 44.11 89.67 45.54 59.77
Madichetty & Sridevi (2021) 75.24 59.81 71.79 68.95 76.19 43.42 77.92 65.84
Proposed 48.39 20.00 96.48 54.96 68.18 17.65 94.36 60.06

Notes.

The bold values are the highest performances achieved by the proposed model for each disaster.