---
language: en
license: mit
datasets:
- twitter-sentiment
metrics:
- accuracy
- f1
- precision
- recall
base_model:
- google-bert/bert-base-uncased
pipeline_tag: text-classification
library_name: transformers
---

# Fine-Tuned BERT Sentiment Model

This model was fine-tuned for sentiment classification.

- Pre-trained model used:  google-bert/bert-base-uncased.
- Dataset used:            twitter-sentiment.
- max_length = 128
- batch_size = 8
- learning_rate = 1e-4
- epochs = 3

## **Evaluation Results**

### 📌 **Before Fine-Tuning**
**Accuracy:** 0.4046

| Class      | Precision | Recall | F1-Score | Support |
|------------|------------|------------|------------|------------|
| Negative   | 0.00       | 0.00       | 0.00       | 1001 |
| Neutral    | 0.40       | 1.00       | 0.58       | 1430 |
| Positive   | 0.00       | 0.00       | 0.00       | 1103 |
| **Macro Avg**  | 0.13   | 0.33       | 0.19       | 3534 |
| **Weighted Avg**  | 0.16   | 0.40   | 0.23   | 3534 |

---

### ✅ **After Fine-Tuning**
**Accuracy:** 0.6095

| Class      | Precision | Recall | F1-Score | Support |
|------------|------------|------------|------------|------------|
| Negative   | 0.82       | 0.29       | 0.42       | 1001 |
| Neutral    | 0.51       | 0.89       | 0.65       | 1430 |
| Positive   | 0.85       | 0.54       | 0.66       | 1103 |
| **Macro Avg**  | 0.73   | 0.57       | 0.58       | 3534 |
| **Weighted Avg**  | 0.70   | 0.61   | 0.59   | 3534 |

---

You can download the model from [Hugging Face](https://huggingface.co/Wolverine001/bert_finetuned_senti).