deprem-ml/deprem_bert_128k_v13_beta

Eval Results

                  precision    recall  f1-score   support
        Lojistik       0.81      0.79      0.80        38
Elektrik Kaynagi       0.73      0.91      0.81        56
  Arama Ekipmani       0.79      0.74      0.76       128
          Cenaze       1.00      0.50      0.67         2
           Giysi       0.82      0.96      0.89       138
  Enkaz Kaldirma       0.94      0.94      0.94       919
          Isinma       0.84      0.89      0.86       185
         Barınma       0.96      0.96      0.96       483
         Tuvalet       0.67      0.80      0.73        10
              Su       0.83      0.87      0.85        67
           Yemek       0.89      0.96      0.92       202
          Saglik       0.80      0.88      0.83       104
        Alakasiz       0.90      0.82      0.86       377

       micro avg       0.89      0.91      0.90      2709
       macro avg       0.84      0.85      0.84      2709
    weighted avg       0.90      0.91      0.90      2709
     samples avg       0.91      0.92      0.91      2709

Threshold:

• Best Threshold: 0.53

Class Loss Weights

    [3.017203135650159,
     2.4823691788825464,
     1.941736822154725,
     6.172646581418988,
     1.8759436445637834,
     1.0,
     1.75011143349181,
     1.2730236191357969,
     4.849237178079731,
     2.4857419672410703,
     1.6324480531290084,
     2.0033774839735035,
     1.3688883733394182]