Skip to content

Latest commit

 

History

History
401 lines (317 loc) · 18.8 KB

训练日志.md

File metadata and controls

401 lines (317 loc) · 18.8 KB

训练日志

FastText

Test Loss: 0.44, Test Acc: 87.18% ECE Loss for epoch 90: 0.06751407352121834 Precision, Recall and F1-Score... precision recall f1-score support

       冒充电商物流客服类     0.8374    0.7387    0.7850      1102
       贷款、代办信用卡类     0.9447    0.9234    0.9339       888
       虚假网络投资理财类     0.8649    0.8795    0.8721       946
        冒充领导、熟人类     0.9177    0.8527    0.8840       353
     冒充公检法及政府机关类     0.9146    0.7918    0.8488       365
     网络游戏产品虚假交易类     0.8962    0.9535    0.9239       172
           刷单返利类     0.9403    0.9725    0.9561      2835
           虚假征信类     0.7079    0.9129    0.7974       677
       冒充军警购物类诈骗     0.7640    0.7727    0.7684        88
        虚假购物、服务类     0.6880    0.6294    0.6574       564
            网黑案件     0.9216    0.9792    0.9495        96

网络婚恋、交友类(非虚假网络投资理财类) 0.6800 0.3923 0.4976 130

        accuracy                         0.8718      8216
       macro avg     0.8398    0.8165    0.8228      8216
    weighted avg     0.8726    0.8718    0.8696      8216

Confusion Matrix... [[ 814 7 17 3 5 4 24 179 0 47 0 2] [ 12 820 4 2 1 0 6 35 0 5 1 2] [ 6 6 832 0 1 1 71 0 0 28 0 1] [ 7 1 6 301 3 0 5 4 1 23 0 2] [ 14 5 2 8 289 0 4 17 8 16 1 1] [ 1 0 2 0 0 164 1 0 0 4 0 0] [ 11 2 35 3 4 0 2757 1 0 15 0 7] [ 36 14 1 1 4 1 1 618 0 1 0 0] [ 1 0 1 2 2 0 0 0 68 14 0 0] [ 68 12 30 5 5 12 36 19 12 355 2 8] [ 0 0 0 0 0 0 1 0 0 0 94 1] [ 2 1 32 3 2 1 26 0 0 8 4 51]]

TextCNN

Test Loss: 1.8, Test Acc: 86.82% Precision, Recall and F1-Score... precision recall f1-score support

       冒充电商物流客服类     0.8716    0.7269    0.7927      1102
       贷款、代办信用卡类     0.9486    0.9358    0.9422       888
       虚假网络投资理财类     0.8556    0.9080    0.8810       946
        冒充领导、熟人类     0.8725    0.8527    0.8625       353
     冒充公检法及政府机关类     0.9174    0.8219    0.8671       365
     网络游戏产品虚假交易类     0.8736    0.9244    0.8983       172
           刷单返利类     0.9457    0.9771    0.9611      2835
           虚假征信类     0.7154    0.9394    0.8123       677
       冒充军警购物类诈骗     0.0000    0.0000    0.0000        88
        虚假购物、服务类     0.5938    0.6791    0.6336       564
            网黑案件     0.9300    0.9688    0.9490        96

网络婚恋、交友类(非虚假网络投资理财类) 0.0000 0.0000 0.0000 130

        accuracy                         0.8682      8216
       macro avg     0.7104    0.7278    0.7166      8216
    weighted avg     0.8514    0.8682    0.8571      8216

Confusion Matrix... [[ 801 6 18 6 4 8 19 178 0 62 0 0] [ 5 831 3 0 0 0 5 41 0 2 1 0] [ 4 4 859 1 1 0 56 0 0 21 0 0] [ 8 2 3 301 5 0 3 3 0 27 1 0] [ 13 4 3 7 300 0 1 14 0 23 0 0] [ 2 1 3 0 0 159 2 0 0 5 0 0] [ 4 3 37 4 2 1 2770 0 0 14 0 0] [ 23 11 1 0 2 1 1 636 0 2 0 0] [ 3 0 1 5 5 0 0 0 0 74 0 0] [ 53 13 39 4 5 10 39 17 0 383 1 0] [ 0 0 0 1 0 0 1 0 0 1 93 0] [ 3 1 37 16 3 3 32 0 0 31 4 0]]

TextRNN

Test Loss: 0.47, Test Acc: 86.39% Precision, Recall and F1-Score... precision recall f1-score support

       冒充电商物流客服类     0.8256    0.7260    0.7726      1102
       贷款、代办信用卡类     0.9612    0.9200    0.9402       888
       虚假网络投资理财类     0.8713    0.8805    0.8759       946
        冒充领导、熟人类     0.8833    0.7932    0.8358       353
     冒充公检法及政府机关类     0.9039    0.8247    0.8625       365
     网络游戏产品虚假交易类     0.8953    0.8953    0.8953       172
           刷单返利类     0.9521    0.9608    0.9565      2835
           虚假征信类     0.6726    0.9498    0.7875       677
       冒充军警购物类诈骗     0.7500    0.7159    0.7326        88
        虚假购物、服务类     0.6556    0.5975    0.6252       564
            网黑案件     0.9574    0.9375    0.9474        96

网络婚恋、交友类(非虚假网络投资理财类) 0.5091 0.4308 0.4667 130

        accuracy                         0.8639      8216
       macro avg     0.8198    0.8027    0.8082      8216
    weighted avg     0.8680    0.8639    0.8632      8216

Confusion Matrix... [[ 800 6 6 2 3 6 20 213 0 46 0 0] [ 9 817 1 1 2 0 6 49 0 2 0 1] [ 18 4 833 1 1 1 58 0 0 17 0 13] [ 10 0 1 280 12 0 0 5 2 32 0 11] [ 10 2 2 6 301 1 1 21 8 11 1 1] [ 4 0 5 0 1 154 0 0 0 8 0 0] [ 6 2 51 1 3 0 2724 0 0 32 0 16] [ 17 9 1 1 2 1 0 643 0 2 0 1] [ 0 0 1 5 1 0 0 0 63 17 0 1] [ 95 10 22 13 7 9 28 25 11 337 1 6] [ 0 0 0 1 0 0 1 0 0 0 90 4] [ 0 0 33 6 0 0 23 0 0 10 2 56]]

TextRNN_Att

Test Loss: 2.3, Test Acc: 34.51% Precision, Recall and F1-Score... precision recall f1-score support

       冒充电商物流客服类     0.0000    0.0000    0.0000      1102
       贷款、代办信用卡类     0.0000    0.0000    0.0000       888
       虚假网络投资理财类     0.0000    0.0000    0.0000       946
        冒充领导、熟人类     0.0000    0.0000    0.0000       353
     冒充公检法及政府机关类     0.0000    0.0000    0.0000       365
     网络游戏产品虚假交易类     0.0000    0.0000    0.0000       172
           刷单返利类     0.3451    1.0000    0.5131      2835
           虚假征信类     0.0000    0.0000    0.0000       677
       冒充军警购物类诈骗     0.0000    0.0000    0.0000        88
        虚假购物、服务类     0.0000    0.0000    0.0000       564
            网黑案件     0.0000    0.0000    0.0000        96

网络婚恋、交友类(非虚假网络投资理财类) 0.0000 0.0000 0.0000 130

        accuracy                         0.3451      8216
       macro avg     0.0288    0.0833    0.0428      8216
    weighted avg     0.1191    0.3451    0.1770      8216

Confusion Matrix... [[ 0 0 0 0 0 0 1102 0 0 0 0 0] [ 0 0 0 0 0 0 888 0 0 0 0 0] [ 0 0 0 0 0 0 946 0 0 0 0 0] [ 0 0 0 0 0 0 353 0 0 0 0 0] [ 0 0 0 0 0 0 365 0 0 0 0 0] [ 0 0 0 0 0 0 172 0 0 0 0 0] [ 0 0 0 0 0 0 2835 0 0 0 0 0] [ 0 0 0 0 0 0 677 0 0 0 0 0] [ 0 0 0 0 0 0 88 0 0 0 0 0] [ 0 0 0 0 0 0 564 0 0 0 0 0] [ 0 0 0 0 0 0 96 0 0 0 0 0] [ 0 0 0 0 0 0 130 0 0 0 0 0]]

28367➗82208 = 34.51%

Bert

Test Loss: 0.43, Test Acc: 86.40% Precision, Recall and F1-Score... precision recall f1-score support

       冒充电商物流客服类     0.9108    0.6579    0.7640      1102
       贷款、代办信用卡类     0.9355    0.9313    0.9334       888
       虚假网络投资理财类     0.8436    0.9006    0.8712       946
        冒充领导、熟人类     0.9314    0.8074    0.8649       353
     冒充公检法及政府机关类     0.9281    0.8137    0.8672       365
     网络游戏产品虚假交易类     0.9191    0.9244    0.9217       172
           刷单返利类     0.9254    0.9802    0.9520      2835
           虚假征信类     0.6774    0.9365    0.7861       677
       冒充军警购物类诈骗     0.7439    0.6932    0.7176        88
        虚假购物、服务类     0.6264    0.6152    0.6208       564
            网黑案件     0.9674    0.9271    0.9468        96

网络婚恋、交友类(非虚假网络投资理财类) 0.7333 0.3385 0.4632 130

        accuracy                         0.8640      8216
       macro avg     0.8452    0.7938    0.8091      8216
    weighted avg     0.8699    0.8640    0.8606      8216

Confusion Matrix... [[ 725 13 22 3 7 5 29 212 0 86 0 0] [ 2 827 2 0 3 0 7 39 0 7 0 1] [ 3 4 852 2 0 0 63 0 0 20 0 2] [ 3 0 9 285 5 0 6 5 2 32 0 6] [ 7 2 5 5 297 0 3 22 8 14 0 2] [ 0 0 4 0 0 159 4 0 0 5 0 0] [ 2 3 37 1 2 1 2779 0 0 9 0 1] [ 15 25 1 0 1 0 1 634 0 0 0 0] [ 0 0 1 1 0 0 0 0 61 25 0 0] [ 39 9 44 7 5 8 66 24 11 347 1 3] [ 0 0 0 0 0 0 5 0 0 1 89 1] [ 0 1 33 2 0 0 40 0 0 8 2 44]]

ERAIN

Test Loss: 0.37, Test Acc: 88.40% Precision, Recall and F1-Score... precision recall f1-score support

       冒充电商物流客服类     0.8582    0.7523    0.8017      1102
       贷款、代办信用卡类     0.9684    0.9313    0.9495       888
       虚假网络投资理财类     0.8627    0.9165    0.8888       946
        冒充领导、熟人类     0.9356    0.8640    0.8984       353
     冒充公检法及政府机关类     0.9379    0.8685    0.9018       365
     网络游戏产品虚假交易类     0.9034    0.9244    0.9138       172
           刷单返利类     0.9506    0.9714    0.9609      2835
           虚假征信类     0.7301    0.9069    0.8090       677
       冒充军警购物类诈骗     0.7901    0.7273    0.7574        88
        虚假购物、服务类     0.6720    0.6720    0.6720       564
            网黑案件     0.9588    0.9688    0.9637        96

网络婚恋、交友类(非虚假网络投资理财类) 0.7746 0.4231 0.5473 130

        accuracy                         0.8840      8216
       macro avg     0.8619    0.8272    0.8387      8216
    weighted avg     0.8861    0.8840    0.8827      8216

Confusion Matrix... [[ 829 3 12 5 5 5 18 162 0 63 0 0] [ 12 827 1 1 1 0 3 34 0 6 1 2] [ 3 3 867 0 1 0 51 0 0 18 0 3] [ 6 2 3 305 5 0 1 2 1 24 0 4] [ 7 1 3 6 317 0 0 13 9 9 0 0] [ 0 0 2 0 0 159 1 0 0 10 0 0] [ 4 3 46 2 0 0 2754 0 0 23 0 3] [ 49 7 1 0 4 1 1 614 0 0 0 0] [ 0 0 1 0 0 0 0 0 64 23 0 0] [ 56 8 37 5 5 11 36 16 7 379 1 3] [ 0 0 0 0 0 0 2 0 0 0 93 1] [ 0 0 32 2 0 0 30 0 0 9 2 55]]

8分类:

FastText:

Test Loss: 0.33, Test Acc: 90.40% Precision, Recall and F1-Score... precision recall f1-score support

           冒充客服类     0.8992    0.9331    0.9159      1779
       贷款、代办信用卡类     0.9517    0.9313    0.9414       888
       虚假网络投资理财类     0.8671    0.8827    0.8748       946
        冒充领导、熟人类     0.9000    0.8414    0.8697       353
     冒充公检法及政府机关类     0.9327    0.7973    0.8597       365
           刷单返利类     0.9329    0.9718    0.9520      2835
         虚假购物服务类     0.7885    0.7512    0.7694       824

网络婚恋、交友类(非虚假网络投资理财类) 0.9045 0.6311 0.7435 225

        accuracy                         0.9040      8215
       macro avg     0.8971    0.8425    0.8658      8215
    weighted avg     0.9034    0.9040    0.9025      8215

Confusion Matrix... [[1660 23 12 2 7 21 54 0] [ 41 827 3 1 1 8 6 1] [ 8 3 835 5 0 75 20 0] [ 5 1 6 297 5 7 30 2] [ 26 3 1 7 291 6 27 4] [ 14 4 40 2 0 2755 16 4] [ 90 8 36 11 8 48 619 4] [ 2 0 30 5 0 33 13 142]]

TextCNN:

Test Loss: 1.4, Test Acc: 91.70% Precision, Recall and F1-Score... precision recall f1-score support

           冒充客服类     0.9150    0.9382    0.9265      1779
       贷款、代办信用卡类     0.9577    0.9426    0.9501       888
       虚假网络投资理财类     0.8818    0.9070    0.8942       946
        冒充领导、熟人类     0.9311    0.8810    0.9054       353
     冒充公检法及政府机关类     0.9450    0.8000    0.8665       365
           刷单返利类     0.9418    0.9764    0.9588      2835
         虚假购物服务类     0.8090    0.7864    0.7975       824

网络婚恋、交友类(非虚假网络投资理财类) 0.9317 0.6667 0.7772 225

        accuracy                         0.9170      8215
       macro avg     0.9141    0.8623    0.8845      8215
    weighted avg     0.9169    0.9170    0.9159      8215

Confusion Matrix... [[1669 17 11 0 5 22 55 0] [ 38 837 3 2 2 3 3 0] [ 7 3 858 0 1 58 18 1] [ 3 0 8 311 0 5 26 0] [ 22 4 2 7 292 4 29 5] [ 8 4 33 2 1 2768 15 4] [ 76 9 31 5 7 47 648 1] [ 1 0 27 7 1 32 7 150]] Time usage: 0:00:00

TextRNN:

Test Loss: 0.37, Test Acc: 89.85% Precision, Recall and F1-Score... precision recall f1-score support

           冒充客服类     0.9065    0.9264    0.9163      1779
       贷款、代办信用卡类     0.9702    0.9178    0.9433       888
       虚假网络投资理财类     0.8686    0.8805    0.8745       946
        冒充领导、熟人类     0.9392    0.7875    0.8567       353
     冒充公检法及政府机关类     0.9201    0.7890    0.8496       365
           刷单返利类     0.9459    0.9623    0.9540      2835
         虚假购物服务类     0.6988    0.7998    0.7459       824

网络婚恋、交友类(非虚假网络投资理财类) 0.8148 0.5867 0.6822 225

        accuracy                         0.8985      8215
       macro avg     0.8830    0.8312    0.8528      8215
    weighted avg     0.9013    0.8985    0.8984      8215

Confusion Matrix... [[1648 11 9 0 4 16 91 0] [ 55 815 3 2 3 3 7 0] [ 4 3 833 0 0 58 43 5] [ 4 0 3 278 6 2 57 3] [ 29 3 1 6 288 6 29 3] [ 8 2 51 2 0 2728 30 14] [ 70 6 29 3 11 41 659 5] [ 0 0 30 5 1 30 27 132]]

TextRNN_Att:

Bert:

Test Loss: 0.34, Test Acc: 90.36% Precision, Recall and F1-Score... precision recall f1-score support

           冒充客服类     0.9123    0.9241    0.9182      1779
       贷款、代办信用卡类     0.9387    0.9313    0.9350       888
       虚假网络投资理财类     0.8452    0.9059    0.8745       946
        冒充领导、熟人类     0.8889    0.8612    0.8748       353
     冒充公检法及政府机关类     0.9406    0.7808    0.8533       365
           刷单返利类     0.9348    0.9714    0.9528      2835
         虚假购物服务类     0.7901    0.7354    0.7618       824

网络婚恋、交友类(非虚假网络投资理财类) 0.9125 0.6489 0.7584 225

        accuracy                         0.9036      8215
       macro avg     0.8954    0.8449    0.8661      8215
    weighted avg     0.9032    0.9036    0.9021      8215

Confusion Matrix... [[1644 24 20 5 7 20 59 0] [ 43 827 3 2 2 5 5 1] [ 3 3 857 2 1 67 13 0] [ 3 0 10 304 1 5 29 1] [ 19 7 5 10 285 5 32 2] [ 7 3 44 3 1 2754 16 7] [ 83 16 45 9 4 58 606 3] [ 0 1 30 7 2 32 7 146]]

ERAIN:

Test Loss: 0.28, Test Acc: 91.89% Precision, Recall and F1-Score... precision recall f1-score support

           冒充客服类     0.9256    0.9376    0.9316      1779
       贷款、代办信用卡类     0.9566    0.9426    0.9495       888
       虚假网络投资理财类     0.8877    0.8943    0.8910       946
        冒充领导、熟人类     0.9472    0.8640    0.9037       353
     冒充公检法及政府机关类     0.9415    0.8384    0.8870       365
           刷单返利类     0.9468    0.9739    0.9602      2835
         虚假购物服务类     0.7879    0.8070    0.7974       824

网络婚恋、交友类(非虚假网络投资理财类) 0.9045 0.7156 0.7990 225

        accuracy                         0.9189      8215
       macro avg     0.9122    0.8717    0.8899      8215
    weighted avg     0.9192    0.9189    0.9184      8215

Confusion Matrix... [[1668 18 8 1 6 17 61 0] [ 38 837 3 1 1 3 5 0] [ 3 3 846 0 0 64 26 4] [ 3 0 3 305 2 5 34 1] [ 18 2 2 7 306 4 24 2] [ 6 4 32 0 1 2761 24 7] [ 66 10 32 5 8 35 665 3] [ 0 1 27 3 1 27 5 161]]