README.md

---
license: apache-2.0
tags:
- generated_from_trainer
metrics:
- accuracy
- f1
- recall
- precision
model-index:
- name: medium-base-News_About_Gold
  results: []
language:
- en
pipeline_tag: text-classification
---

# medium-base-News_About_Gold

This model is a fine-tuned version of [funnel-transformer/medium-base](https://huggingface.co/funnel-transformer/medium-base).
It achieves the following results on the evaluation set:
- Loss: 0.2838
- Accuracy: 0.9172
- Weighted f1: 0.9170
- Micro f1: 0.9172
- Macro f1: 0.8854
- Weighted recall: 0.9172
- Micro recall: 0.9172
- Macro recall: 0.8859
- Weighted precision: 0.9171
- Micro precision: 0.9172
- Macro precision: 0.8853

## Model description

For more information on how it was created, check out the following link: https://github.com/DunnBC22/NLP_Projects/blob/main/Sentiment%20Analysis/Sentiment%20Analysis%20of%20Commodity%20News%20-%20Gold%20(Transformer%20Comparison)/News%20About%20Gold%20-%20Sentiment%20Analysis%20-%20Funnel%20with%20W%26B.ipynb

This project is part of a comparison of seven (7) transformers. Here is the README page for the comparison: https://github.com/DunnBC22/NLP_Projects/tree/main/Sentiment%20Analysis/Sentiment%20Analysis%20of%20Commodity%20News%20-%20Gold%20(Transformer%20Comparison)

## Intended uses & limitations

This model is intended to demonstrate my ability to solve a complex problem using technology.

## Training and evaluation data

Dataset Source: https://www.kaggle.com/datasets/ankurzing/sentiment-analysis-in-commodity-market-gold

_Input Word Length:_

![Length of Input Text (in Words)](https://github.com/DunnBC22/NLP_Projects/raw/main/Sentiment%20Analysis/Sentiment%20Analysis%20of%20Commodity%20News%20-%20Gold%20(Transformer%20Comparison)/Images/Input%20Word%20Length.png)

_Class Distribution:_

![Length of Input Text (in Words)](https://github.com/DunnBC22/NLP_Projects/raw/main/Sentiment%20Analysis/Sentiment%20Analysis%20of%20Commodity%20News%20-%20Gold%20(Transformer%20Comparison)/Images/Class%20Distribution.png)

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy | Weighted F1 | Micro F1 | Macro F1 | Weighted Recall | Micro Recall | Macro Recall | Weighted Precision | Micro Precision | Macro Precision |
|:-------------:|:-----:|:----:|:---------------:|:--------:|:-----------:|:--------:|:--------:|:---------------:|:------------:|:------------:|:------------------:|:---------------:|:---------------:|
| 0.7426        | 1.0   | 133  | 0.3820          | 0.8803   | 0.8636      | 0.8803   | 0.6690   | 0.8803          | 0.8803       | 0.6809       | 0.8862             | 0.8803          | 0.8992          |
| 0.332         | 2.0   | 266  | 0.3083          | 0.9007   | 0.8987      | 0.9007   | 0.8525   | 0.9007          | 0.9007       | 0.8402       | 0.9015             | 0.9007          | 0.8705          |
| 0.2381        | 3.0   | 399  | 0.2870          | 0.9106   | 0.9097      | 0.9106   | 0.8686   | 0.9106          | 0.9106       | 0.8539       | 0.9096             | 0.9106          | 0.8862          |
| 0.1911        | 4.0   | 532  | 0.2797          | 0.9163   | 0.9158      | 0.9163   | 0.8843   | 0.9163          | 0.9163       | 0.8819       | 0.9159             | 0.9163          | 0.8873          |
| 0.1584        | 5.0   | 665  | 0.2838          | 0.9172   | 0.9170      | 0.9172   | 0.8854   | 0.9172          | 0.9172       | 0.8859       | 0.9171             | 0.9172          | 0.8853          |


### Framework versions

- Transformers 4.28.1
- Pytorch 2.0.0
- Datasets 2.11.0
- Tokenizers 0.13.3