Developer's note
Please do download and try out the model locally or on colab, as it helps huggingface determine that this model is important enough to have a serverless API for everyone to use. Also, the model is totally safe for everyone to use. The only reason one of the files has been marked unsafe because it is a pickle file.Thank you all for so much support !!
Overview
This model is a fine-tuned version of google-bert/bert-base-uncased on this Kaggle dataset. It achieves the following results on the evaluation set:
- Macro f1: 89.44%
- Weighted f1: 93.15%
- Accuracy: 93.80%
- Balanced accuracy: 90.42%
Model description
This finetuned version of google-bert/bert-base-uncased excels at detecting the crime type from the description of the crime. It has 34 labels.
Training and evaluation data
- eval_macro f1: 89.44%
- eval_weighted f1: 93.15%
- eval_accuracy: 93.79%
- eval_balanced accuracy: 90.42%
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 30
Training results
Training Loss | Epoch | Step | Validation Loss | Macro f1 | Weighted f1 | Accuracy | Balanced accuracy |
---|---|---|---|---|---|---|---|
0.1859 | 1.0 | 5538 | 0.1297 | 0.8561 | 0.9249 | 0.9366 | 0.8571 |
0.1281 | 2.0 | 11076 | 0.1260 | 0.8702 | 0.9248 | 0.9369 | 0.8740 |
0.1279 | 3.0 | 16614 | 0.1251 | 0.8728 | 0.9314 | 0.9380 | 0.8749 |
0.1272 | 4.0 | 22152 | 0.1276 | 0.8652 | 0.9247 | 0.9367 | 0.8655 |
0.1266 | 5.0 | 27690 | 0.1256 | 0.8685 | 0.9252 | 0.9345 | 0.8724 |
0.1284 | 6.0 | 33228 | 0.1264 | 0.8668 | 0.9252 | 0.9345 | 0.8724 |
0.1272 | 7.0 | 38766 | 0.1247 | 0.8739 | 0.9313 | 0.9379 | 0.8748 |
0.1262 | 8.0 | 44304 | 0.1258 | 0.8892 | 0.9246 | 0.9366 | 0.9024 |
0.1263 | 9.0 | 49842 | 0.1251 | 0.9038 | 0.9310 | 0.9378 | 0.9041 |
0.1267 | 10.0 | 55380 | 0.1244 | 0.8897 | 0.9253 | 0.9345 | 0.9018 |
0.1271 | 11.0 | 60918 | 0.1251 | 0.8951 | 0.9325 | 0.9371 | 0.9036 |
0.1268 | 12.0 | 66456 | 0.1248 | 0.8944 | 0.9315 | 0.9380 | 0.9042 |
0.1254 | 13.0 | 71994 | 0.1247 | 0.9038 | 0.9314 | 0.9381 | 0.9043 |
0.126 | 14.0 | 77532 | 0.1263 | 0.8944 | 0.9314 | 0.9379 | 0.9042 |
0.1261 | 15.0 | 83070 | 0.1274 | 0.8891 | 0.9250 | 0.9348 | 0.9020 |
0.1253 | 16.0 | 88608 | 0.1241 | 0.8944 | 0.9315 | 0.9380 | 0.9042 |
0.1251 | 17.0 | 94146 | 0.1244 | 0.9042 | 0.9314 | 0.9380 | 0.9042 |
0.125 | 18.0 | 99684 | 0.1249 | 0.9041 | 0.9314 | 0.9380 | 0.9043 |
0.125 | 19.0 | 105222 | 0.1245 | 0.8942 | 0.9312 | 0.9380 | 0.9042 |
0.1257 | 20.0 | 110760 | 0.1248 | 0.9041 | 0.9313 | 0.9379 | 0.9042 |
0.125 | 21.0 | 116298 | 0.1248 | 0.9000 | 0.9254 | 0.9344 | 0.9018 |
0.1248 | 22.0 | 121836 | 0.1244 | 0.9041 | 0.9313 | 0.9379 | 0.9042 |
0.1246 | 23.0 | 127374 | 0.1245 | 0.9042 | 0.9315 | 0.9380 | 0.9042 |
0.1247 | 24.0 | 132912 | 0.1242 | 0.8943 | 0.9314 | 0.9380 | 0.9043 |
0.1245 | 25.0 | 138450 | 0.1242 | 0.9042 | 0.9315 | 0.9380 | 0.9042 |
0.1245 | 26.0 | 143988 | 0.1245 | 0.9042 | 0.9314 | 0.9381 | 0.9043 |
0.1245 | 27.0 | 149526 | 0.1242 | 0.8944 | 0.9314 | 0.9381 | 0.9043 |
0.1244 | 28.0 | 155064 | 0.1242 | 0.9336 | 0.9315 | 0.9381 | 0.9337 |
0.1243 | 29.0 | 160602 | 0.1243 | 0.8944 | 0.9314 | 0.9381 | 0.9043 |
0.1243 | 30.0 | 166140 | 0.1243 | 0.8944 | 0.9314 | 0.9381 | 0.9043 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.2.2
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 38
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for Luna-Skywalker/BERT-crime-analysis
Base model
google-bert/bert-base-uncased