Explore adding market sentiment from social media (or other sources?)
One idea:
- use a reddit crawler to get largest bitcoin/eth subreddits from 2018 onward, save each comments timestamp
- push it through a bert-based sentiment analyser
- aggregate timestamp-sentiment pairs to minute level so we can use it at bar creation
- if there are missing sentiment values for a given minute bar, we could fill in the nan with the average sentiment?
Explore adding market sentiment from social media (or other sources?)
One idea: