In order to run the code initially, first you need to get the model in the link (https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment-latest) and later you need to visit https://www.kaggle.com/datasets/yash612/stockmarket-sentiment-dataset/data to get the sentiment fine-tuning dataset, when you have completed the above steps you can run pretrain.py, the code will save the fine-tuned model after running, and next you can run eval.py to evaluate the fine-tuning Next you can run eval.py to evaluate the effect of the fine-tuned model, which is usually not too different from the paper.
To complete the stock price-sentiment correlation experiment, you need to visit the link (https://www.kaggle.com/code/equinxx/stock-prediction-gan-twitter-sentiment-analysis/input) to get the paper's reference to the The second dataset, followed by running the predict.ipynb file is sufficient. If you want to get the AAPL results from the paper, you need to visit the link (https://www.kaggle.com/datasets/omermetinn/tweets-about-the-top-companies-from-2015-to-2020/data?select=Company.csv) and (https://www.kaggle.com/datasets/omermetinn/values-of-top-nasdaq-copanies-from-2010-to-2020) and subsequently run the predict2.ipynb code.
Regarding the latter two experiments, you can see the contents directly in the ipynb file without trying to run it.
The paper on this experiment is under review, and I'll post a link to it subsequently
If you have any questions feel free to contact me, and if you find this job helpful, I hope to get your STAR!