sec-sentiment-pred

In this repository, we create a bag-of-words model that predicts whether 8-K documents forecast rise or fall in stock price

Dependencies

We use EDGAR-DOC-PARSER to create a dataset.

Conda

To create conda environment:

conda create -n sectagging -c conda-forge -c anaconda python=3.10 pytorch scipy numpy selenium=4.5.0 pyyaml chardet requests lxml scikit-learn pandas pytorch
conda activate sectagging
pip install edgar-doc-parser==0.2.2.post1

Pipeline

Use generate_data.py to make 8k_data_raw.tsv Use filter.py on 8k_data_raw.tsv to make 8k_data_filtered.tsv Use dataloading.py on 8k_data_filtered.tsv to make 8k_data.tsv, the final version

For test runs, use generate_short.py on 8k_data.tsv to make 8k_data_short.tsv