fnet-base-Financial_Sentiment_Analysis

This model is a fine-tuned version of google/fnet-base.

It achieves the following results on the evaluation set:

• Loss: 0.3281
• Accuracy: 0.8117
• Weighted f1: 0.8110
• Micro f1: 0.8117
• Macro f1: 0.7472
• Weighted recall: 0.8117
• Micro recall: 0.8117
• Macro recall: 0.7394
• Weighted precision: 0.8144
• Micro precision: 0.8117
• Macro precision: 0.7588

Model description

This is a sentiment analysis (text classification) model of statements about finances.

For more information on how it was created, check out the following link: https://github.com/DunnBC22/NLP_Projects/blob/main/Sentiment%20Analysis/Financial%20Sentiment%20Analysis/Financial_Sentiment_Analysis_v2.ipynb

Intended uses & limitations

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

Training and evaluation data

Dataset Sources:

• https://www.kaggle.com/datasets/sbhatti/financial-sentiment-analysis
• https://www.kaggle.com/datasets/ankurzing/sentiment-analysis-for-financial-news

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
0.6116	1.0	134	0.5127	0.6304	0.5606	0.6304	0.4705	0.6304	0.6304	0.5272	0.6722	0.6304	0.6103
0.4497	2.0	268	0.3885	0.7578	0.7490	0.7578	0.6783	0.7578	0.7578	0.6636	0.7677	0.7578	0.7196
0.3319	3.0	402	0.3546	0.7799	0.7784	0.7799	0.7185	0.7799	0.7799	0.7167	0.7979	0.7799	0.7440
0.2953	4.0	536	0.3312	0.8117	0.8105	0.8117	0.7435	0.8117	0.8117	0.7356	0.8111	0.8117	0.7532
0.2446	5.0	670	0.3281	0.8117	0.8110	0.8117	0.7472	0.8117	0.8117	0.7394	0.8144	0.8117	0.7588

Framework versions

• Transformers 4.27.4
• Pytorch 2.0.0
• Datasets 2.11.0
• Tokenizers 0.13.3