Analyzing Sentiment and Recent Sales Data to Understand the Most Promising NFT Projects
Nonfungible tokens, or NFTs, are one of a kind digital property. While the digital art market has been on fire with many items selling for millions of dollars (US), the technology and role NFTs will play in culture is so new we are still trying, collectively, to understand their value and use cases.
For investors, it would be useful to place a more (though never totally) objective framework around NFT art projects to discern where they should be focusing their attention and capital. Art is, of course, subjective and perhaps others will be willing to purchase a project simply because they enjoy the art's aestetics. However, for others, they want to purchase the NFT because while they don't dislike the art, it is a well-known or popularly rising project, and they believe it will increase in value over a time frame.
Twitter seems to be as good of a source of mood as any regarding which projects are waning and waxing in popularity and potential. We built a sentiment analyzer of tweets containing the name of a list of top projects, according to the Rarity.tools website. The goal is to use the sentiment data (along with price data to determine if you can afford to buy in) to find the projects have the highest sentiment scores in concert with strong sales data. The idea is NOT that this is a "You Should Buy This: xyz" project but rather a guide to show which projects have strong positive feelings surrounding them.
The next version would attempt to examine the newest projects with the hopes of identifying projects with the largest early hype as an investment strategy. The space is so young there is a lot more research that needs to be done to undertand the trajectory of new and established projects.
Finally, we also attempted a price prediction model on crypto punks because the project has been around for a while and there is plenty of price data avaialble. The idea is developing a model and being able to extrapolate the model to other projects with less data in hopes of predicting the price. The results of prediction were not good enough to use in production. Further refinement is necessary.