---
tags:
- generated_from_trainer
- finance
base_model: cardiffnlp/twitter-roberta-base-sentiment
metrics:
- accuracy
model-index:
- name: fine-tuned-cardiffnlp-twitter-roberta-base-sentiment-finance-dataset
  results: []
datasets:
- CJCJ3030/twitter-financial-news-sentiment
language:
- en
library_name: transformers
pipeline_tag: text-classification
widget:
- text: "UK house sales up 12% in April"
- text: "Singapore oil trader convicted of abetting forgery and cheating HSBC"
- text: "‘There’s money everywhere’: Milken conference-goers look for a dealmaking revival"
- text: "ETF buying nearly halves in April as US rate cut hopes recede"
- text: "Todd Boehly’s investment house in advanced talks to buy private credit firm"
- text: "Berkshire Hathaway’s cash pile hits new record as Buffett dumps stocks"
- text: "Harvest partnership to bring HK-listed crypto ETFs to Singapore"
- text: "Kazakh oligarch Timur Kulibayev sells Mayfair mansion for £35mn"
- text: "Deutsche Bank’s DWS inflated client asset inflows by billions of euro"
- text: "UBS reports stronger than expected profit in first quarter"
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# fine-tuned-cardiffnlp-twitter-roberta-base-sentiment-finance-dataset

This model is a fine-tuned version of [cardiffnlp/twitter-roberta-base-sentiment](https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment) on an twitter finance news sentiment dataset.
It achieves the following results on the evaluation set:
- Loss: 0.3123
- Accuracy: 0.8559

10 examples in Inference API are gathered from https://twitter.com/ftfinancenews in early may 2024

Colab Notebook for fine tuning : https://colab.research.google.com/drive/1gvpFbazlxg3AdSldH3w6TYjGUByxqCrh?usp=sharing

### Training Data

https://huggingface.co/datasets/CJCJ3030/twitter-financial-news-sentiment/viewer/default/train

### Evaluation Data

https://huggingface.co/datasets/CJCJ3030/twitter-financial-news-sentiment/viewer/default/validation

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 120
- eval_batch_size: 120
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5

### Training results

| Epoch | Step | Validation Loss | Accuracy |
|:-----:|:----:|:---------------:|:--------:|
| 1.0   | 80   | 0.3123          | 0.8559   |
| 2.0   | 160  | 0.3200          | 0.8576   |
| 3.0   | 240  | 0.3538          | 0.8819   |
| 4.0   | 320  | 0.3695          | 0.8882   |
| 5.0   | 400  | 0.4108          | 0.8869   |


### Framework versions

- Transformers 4.40.2
- Pytorch 2.2.1+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1

## Citation

```bibtex
@inproceedings{barbieri-etal-2020-tweeteval,
    title = "{T}weet{E}val: Unified Benchmark and Comparative Evaluation for Tweet Classification",
    author = "Barbieri, Francesco  and
      Camacho-Collados, Jose  and
      Espinosa Anke, Luis  and
      Neves, Leonardo",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2020",
    month = nov,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.findings-emnlp.148",
    doi = "10.18653/v1/2020.findings-emnlp.148",
    pages = "1644--1650"
}
```