---
language: en
license: apache-2.0
---

# Women's Clothing Reviews Sentiment Analysis with DistilBERT

## Overview

This Hugging Face repository contains a fine-tuned DistilBERT model for sentiment analysis of women's clothing reviews. The model is designed to classify reviews into positive, negative, or neutral sentiment categories, providing valuable insights into customer opinions.

## Model Details

- **Model Architecture**: Fine-tuned DistilBERT
- **Sentiment Categories**: Neutral [0], Negative [1], Positive [2]
- **Input Format**: Text-based clothing reviews
- **Output Format**: Sentiment category labels

## Fine-tuning procedure
This model was fine-tuned using a relatively small dataset containing 23487 rows broken down into train/eval/test dataset. Nevertheless, the fine-tuned model was able to performs slightly better than the base-distilbert-model on the test dataset.


## Training result
It achieved the following results on the evaluation set:
- **Validation Loss**: 1.1677

### Comparison between the base distilbert model VS fine-tuned distilbert
| Model          | Accuracy | Precision | Recall | F1 Score |
|--------------- | -------- | --------- | ------ | -------- |
| DistilBERT base model    | 0.79     | 0.77      | 0.79   | 0.77     |
| DistilBERT fine-tuned     | 0.85     | 0.86      | 0.85   | 0.85     |

 
## Installation

To use this model, you'll need to install the Hugging Face Transformers library and any additional dependencies.
- **pip install transformers**
- **pip install torch**


## Usage
You can easily load the pre-trained model for sentiment analysis using Hugging Face's DistilBertForSequenceClassification and DistilBertTokenizerFast.

```python
from transformers import DistilBertForSequenceClassification, DistilBertTokenizerFast
import torch

model_name = "ongaunjie/distilbert-cloths-sentiment" 
tokenizer = DistilBertTokenizerFast.from_pretrained(model_name)
model = DistilBertForSequenceClassification.from_pretrained(model_name)

review = "This dress is amazing, I love it!"
inputs = tokenizer.encode(review, return_tensors="pt")
with torch.no_grad():
    outputs = model(inputs)
predicted_class = int(torch.argmax(outputs.logits))