Rouse Award: Investigating Bitcoin price prediction
Luca (Upper Sixth) investigated the predictability of Bitcoin prices based on posts on social media platform X for his Rouse Award-winning project.
Considered to be the original cryptocurrency, having been conceived in 2008, the worth of Bitcoin against the US dollar has fluctuated widely.
Using machine learning, Luca carried out a sentiment analysis of posts on X, formerly Twitter, to determine what impact positive or negative views about Bitcoin had on its price over a three-month period.
He said: “Although you can find out about Bitcoin in other ways, there’s still a lot of communication done through Twitter, which made me think there must be some sort of correlation because people are obviously influenced by it.
“I used a pre-trained small language model to carry out the sentiment analysis. The general idea was a tweet had to have at least 250 likes and had to have enough people viewing it. Tweets also had to have been tagged Bitcoin because I was clear I wanted specific data.
“The way I predicted it was using a neural network, so I was feeding other data in as well, so it’s hard to specifically say a positive view equals prices go up, it’s usually a more complex correlation.
“However, I was able to fairly accurately predict whether the price was going to go up or down. I made a few models and slightly tinkered with things, but in the three-month period I tested it over, the value of Bitcoin went up by 50%.”
With a keen interest in the future of artificial intelligence, Luca found the project extremely rewarding.
He said: “I had a general idea about a lot of machine learning because it was something I’d discovered before, but with this project, I was able to test my understanding of different concepts and implement some ideas I had.
“It was just a good way of turning theory into practice, so I really enjoyed doing it.
“I’m applying to study computer science at university, and I’m interested in AI more generally, but I think if you can apply it to economic situations such as this, it’s a valuable skill.”
Read Luca’s project in full at our Perse Research website.