---
license: cc-by-sa-4.0
tags:
- financial-sentiment-analysis
- sentiment-analysis
- sentence_50agree
- generated_from_trainer
- sentiment
- finance
datasets:
- financial_phrasebank
- Kaggle_Self_label
- nickmuchi/financial-classification
metrics:
- accuracy
- f1
- precision
- recall
widget:
- text: The USD rallied by 10% last night
  example_title: Bullish Sentiment
- text: >-
    Covid-19 cases have been increasing over the past few months impacting
    earnings for global firms
  example_title: Bearish Sentiment
- text: the USD has been trending lower
  example_title: Mildly Bearish Sentiment
model-index:
- name: sec-bert-finetuned-finance-classification
  results:
  - task:
      name: Text Classification
      type: text-classification
    dataset:
      name: financial_phrasebank
      type: finance
      args: sentence_50agree
    metrics:
    - type: F1
      name: F1
      value: 0.8744
    - type: accuracy
      name: accuracy
      value: 0.8755
language:
- en
---

<!-- 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. -->

# sec-bert-finetuned-finance-classification

This model is a fine-tuned version of [nlpaueb/sec-bert-base](https://huggingface.co/nlpaueb/sec-bert-base) on the sentence_50Agree [financial-phrasebank + Kaggle Dataset](https://huggingface.co/datasets/nickmuchi/financial-classification), a dataset consisting of 4840 Financial News categorised by sentiment (negative, neutral, positive). The Kaggle dataset includes Covid-19 sentiment data and can be found here: [sentiment-classification-selflabel-dataset](https://www.kaggle.com/percyzheng/sentiment-classification-selflabel-dataset).

It achieves the following results on the evaluation set:
- Loss: 0.5277
- Accuracy: 0.8755
- F1: 0.8744
- Precision: 0.8754
- Recall: 0.8755

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## 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: 8
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1     | Precision | Recall |
|:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|:---------:|:------:|
| 0.6005        | 0.99  | 71   | 0.3702          | 0.8478   | 0.8465 | 0.8491    | 0.8478 |
| 0.3226        | 1.97  | 142  | 0.3172          | 0.8834   | 0.8822 | 0.8861    | 0.8834 |
| 0.2299        | 2.96  | 213  | 0.3313          | 0.8814   | 0.8805 | 0.8821    | 0.8814 |
| 0.1277        | 3.94  | 284  | 0.3925          | 0.8775   | 0.8771 | 0.8770    | 0.8775 |
| 0.0764        | 4.93  | 355  | 0.4517          | 0.8715   | 0.8704 | 0.8717    | 0.8715 |
| 0.0533        | 5.92  | 426  | 0.4851          | 0.8735   | 0.8728 | 0.8731    | 0.8735 |
| 0.0363        | 6.9   | 497  | 0.5107          | 0.8755   | 0.8743 | 0.8757    | 0.8755 |
| 0.0248        | 7.89  | 568  | 0.5277          | 0.8755   | 0.8744 | 0.8754    | 0.8755 |


### Framework versions

- Transformers 4.17.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.4
- Tokenizers 0.11.6