https://github.com/d3n7/StockGPT

Predict Stock movements with GPT-4 or GPT-3.5
https://github.com/d3n7/StockGPT

Last synced: 3 months ago
JSON representation

Predict Stock movements with GPT-4 or GPT-3.5

• Host: GitHub
• URL: https://github.com/d3n7/StockGPT
• Owner: d3n7
• Created: 2023-04-22T02:14:57.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-04-24T03:08:59.000Z (over 1 year ago)
• Last Synced: 2024-06-23T11:36:50.207Z (5 months ago)
• Language: Python
• Size: 35.2 KB
• Stars: 117
• Watchers: 3
• Forks: 21
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-gpt4 - StockGPT - Primitive technique to predict stock movements with GPT-4 or 3.5. (Tools / Open-source projects)

README

        
# StockGPT

Predict Stock movements with GPT-4 or GPT-3.5

inspired by this paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4412788

### Currently Untested!

# DISCLAIMER

Please only use this to see how it correlates to stock movements rather than use it to invest.

it's probably not very good at it anyways.

# How it works

For every company you put in company.txt, it'll search the internet for news articles in from last day about your company.

It will evaluate each headline as a number between -1 and 1, signifying if it will impact the stock price negatively, postively or not at all.

It generates the reports for you on each company's healdines, in the form of a .csv file. And a final report which includes the api costs.

# How to use it

1. If you haven't installed the dependencies already, run, ```pip install -r requirements.txt```

2. If you haven't already, put your OpenAI api token in the file called auth.txt

3. Put a list of companies you want to track in companies.txt

4. Run ```python sgpt.py -h``` to see your options, then run the command as you want.

# Examples

```python sgpt.py -t -c``` gpt-3.5-turbo, sending the headlines in a batch (for minimum cost)





Stock movements the next day:



# Full options

```optional arguments:

  -h, --help            show this help message and exit

  -t, --turbo           use gpt-3.5-turbo instead of gpt-4

  -c, --combined        send and receive all the headlines in bulk (cheaper but probabaly less good)

  -T TEMP, --temp TEMP  temperature (variability) of the model. a value between 0.0 and 1.0 (default: 0.3)```