Coin'ed

Trading through sentiment analysis of scrapped data

Comment

Inspiration

We had a friend who once did something very similar by manually looking at reddit threads to determine which crypto coins to buy. We decided to automate it.

What it does

It scrapes data from reddit using browser.cash, then performs sentiment analysis using a home-trained machine learning model. Then, we used the data and performed another round of fancy mathematics to obtain numbers on how much hype there is around the crypto-coin. Finally, we built an api which we used to automatically buy and sell the crypto-coins which we predicted to either rise in value or decrease in value.

How we built it

With love and the power of friendship, but we actually used brower.cash as a service to bypass verification. We wrote the scraping code using python playwrite, which we fed into browser.cash. After that, we used python again to run the sentiment analysis; as well as the next round of refinement. The refinement round outputted a json data file which our angular front end used to display it. Finally, it connects to the python api to automatically buy and sell crypto currency to well automatically do it.

Challenges we ran into

The integration from the frontend and the backend was very difficult for us. As we built our frontend in angular which uses typescript, while the backend uses python. This caused many difficulties since we originally built our frontend with testdata.

Accomplishments that we're proud of

Using angular instead of react, since for every other hackathon that we've participated in, we've used react to create our frontend.

What we learned

How to use angular, as well as how to use websockets for live communication between our python backend and typescript frontend

What's next for Coin'ed

Expand the source where we get information from. Instead of just scrapping information from reddit, we also aim to scrape information for other sites like x/twitter, bluesky, etc.

Built With

Share this project:

Updates