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

Table 5. Results of best BERT classifiers on the test part of the seven disasters (damage vs non-damage).

Disasters Sequence length TP FN TN FP Accuracy Macro-precision Macro-recall Macro-f1-score
Iraq–Iran earthquake 64 67 2 27 3 0.9495 0.9441 0.9355 0.9396
128 68 1 27 3 0.9596 0.9610 0.9428 0.9512
Sri Lanka floods 64 140 11 14 0 0.9333 0.7800 0.9636 0.8401
128 141 10 14 0 0.9394 0.7917 0.9669 0.8513
Mexico earthquake 64 180 9 39 8 0.9280 0.8850 0.8911 0.8880
128 179 10 38 9 0.9195 0.8719 0.8778 0.8748
California wildfires 64 243 31 52 16 0.8626 0.7824 0.8258 0.8003
128 242 32 49 19 0.8509 0.7661 0.8019 0.7812
Hurricane Harvey 64 631 52 77 34 0.8917 0.7729 0.8088 0.7889
128 636 47 74 37 0.8942 0.7783 0.7989 0.7880
Hurricane Maria 64 679 43 55 19 0.9221 0.7670 0.8418 0.7979
128 670 52 55 19 0.9108 0.7432 0.8356 0.7787
Hurricane Irma 64 658 31 76 35 0.9175 0.8299 0.8198 0.8247
128 658 31 75 36 0.9163 0.8278 0.8153 0.8214