Bert Uncased Model Fine Tuned For Stock Sentiment

  • This model is a fine-tuned version of the BERT (Bidirectional Encoder Representations from Transformers) model specifically designed for analyzing stock sentiment. The fine-tuning process involved training the model on tagged comments from the last two pages of the stock form on the Investing platform, focusing on stocks listed in the BIST Index.

Stock List:

  • ACSEL, ADEL, ARCLK, ASELS, AZTEK, BIMAS, BFREN, BMSCH,
  • CCOLA, CIMSA, CMBTN, CWENE,EKGYO, ENJSA, EREGL, FROTO,
  • GOODY, GUBRF, HALKB, HEKTS, ISCTR, KCHOL, KOZAL, KOPOL,
  • KRDMD, ONCSM, PETKM, PKART, SAHOL, SASA, SISE, SMRTG,
  • THYAO, TMSN, TCELL, TTKOM, TOASO, TTRAK, TUPRS, VESTL, YAPRK, YKSLN

This fine-tuned model aims to provide insights into the sentiment of these stocks based on the given tagged comments and can be used for stock sentiment analysis in financial applications.

Colab File

Training hyperparameters

Training Hyperparameters: The following hyperparameters were used during training:

  • Optimizer: SGD
  • Learning Rate: 3e-2
  • Number of Training Epochs: 10
  • Metric for Best Model: F1 Score

Training Results

Epoch Training Loss Validation Loss Accuracy Precision Recall F1 Score
1 1.057400 0.895725 0.621538 0.618631 0.612559 0.611949
2 0.908400 0.822652 0.632308 0.644781 0.629953 0.622661
3 0.812100 0.788586 0.656923 0.680735 0.659374 0.650310
4 0.747700 0.737312 0.667692 0.670311 0.668073 0.666547
5 0.712600 0.743018 0.692308 0.710226 0.691384 0.686578
6 0.659200 0.771312 0.670769 0.695524 0.669198 0.662246
7 0.608300 0.733821 0.680000 0.677778 0.678871 0.677992
8 0.575900 0.739905 0.701538 0.702704 0.700902 0.698514
9 0.565200 0.754889 0.692308 0.692446 0.693058 0.691157
10 0.541000 0.754683 0.704615 0.705291 0.704209 0.702093

Evaluation Results

Loss Accuracy Precision Recall F1 Score Runtime Samples/s Steps/s Epoch
0.754683 0.704615 0.705291 0.704209 0.702093 3.3869 191.915 24.211 10.0

Framework versions

  • Transformers 4.30.2
  • TensorFlow 2.12.0
  • Tokenizers 0.13.3
Downloads last month
48
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.