Hugging Face's logo

deprem-ml
/
multilabel_earthquake_tweet_intent_bert_base_turkish_cased

Text Classification
Transformers
PyTorch
Turkish
bert
deprem-clf-v1
Eval Results
Inference Endpoints
Model card Files Community
3
nlztrk commited on
Commit
c88955f
•
1 Parent(s): 6ba4f3a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +33 -21
README.md CHANGED
@@ -40,36 +40,48 @@ train_test_split(df_train,
40
  random_state=1111)
41
  ```
42
 
43
- ## Training Log
 
 
 
 
 
 
 
 
 
 
 
 
44
  ```
45
  Epoch Training Loss Validation Loss
46
- 1 No log 0.150276
47
- 2 0.195100 0.132906
48
- 3 0.107700 0.128633
49
- 4 0.107700 0.127795
50
  ```
51
 
52
  ## Threshold Optimization
53
- - **Best Threshold:** 0.1
54
- - **F1 @ Threshold:** 0.734
55
 
56
  ## Eval Results
57
  ```
58
  precision recall f1-score support
59
 
60
- Alakasiz 0.90 0.87 0.89 734
61
- Barinma 0.85 0.80 0.83 207
62
- Elektronik 0.73 0.78 0.75 130
63
- Giysi 0.83 0.66 0.73 94
64
- Kurtarma 0.86 0.79 0.82 362
65
- Lojistik 0.73 0.51 0.60 112
66
- Saglik 0.74 0.74 0.74 108
67
- Su 0.64 0.60 0.62 78
68
- Yagma 0.68 0.55 0.61 31
69
- Yemek 0.80 0.83 0.81 117
70
 
71
- micro avg 0.84 0.79 0.81 1973
72
- macro avg 0.78 0.71 0.74 1973
73
- weighted avg 0.84 0.79 0.81 1973
74
- samples avg 0.84 0.82 0.82 1973
75
  ```
 
40
  random_state=1111)
41
  ```
42
 
43
+ ## Class Loss Weights
44
+ - **Alakasiz:** 1.0
45
+ - **Barinma:** 1.5167249178108022
46
+ - **Elektronik:** 1.7547338578655642
47
+ - **Giysi:** 1.9610520059358458
48
+ - **Kurtarma:** 1.269341370129623
49
+ - **Lojistik:** 1.8684086209021484
50
+ - **Saglik:** 1.8019018017117145
51
+ - **Su:** 2.110648663094536
52
+ - **Yagma:** 3.081208739200435
53
+ - **Yemek:** 1.7994815143101963
54
+
55
+ ## Training Log (Class-Scaled)
56
  ```
57
  Epoch Training Loss Validation Loss
58
+ 1 No log 0.216295
59
+ 2 0.260000 0.171498
60
+ 3 0.142700 0.175608
61
+ 4 0.142700 0.169851
62
  ```
63
 
64
  ## Threshold Optimization
65
+ - **Best Threshold:** 0.15
66
+ - **F1 @ Threshold:** 0.7503
67
 
68
  ## Eval Results
69
  ```
70
  precision recall f1-score support
71
 
72
+ Alakasiz 0.91 0.87 0.89 734
73
+ Barinma 0.85 0.81 0.83 207
74
+ Elektronik 0.72 0.78 0.75 130
75
+ Giysi 0.73 0.67 0.70 94
76
+ Kurtarma 0.86 0.81 0.83 362
77
+ Lojistik 0.68 0.56 0.62 112
78
+ Saglik 0.72 0.81 0.76 108
79
+ Su 0.61 0.69 0.65 78
80
+ Yagma 0.67 0.65 0.66 31
81
+ Yemek 0.79 0.85 0.82 117
82
 
83
+ micro avg 0.82 0.81 0.81 1973
84
+ macro avg 0.75 0.75 0.75 1973
85
+ weighted avg 0.83 0.81 0.81 1973
86
+ samples avg 0.84 0.84 0.83 1973
87
  ```