Using R to Examine Fluctuations in the Price of NEO

Hi friends,

So for work, I’ve been doing a lot of data science stuff with R. On the side, I’ve been playing around with R and NEO, looking for potential underlying drivers of the price of NEO. In essence, what are some of the asset correlates that have been driving the wild swings in the price of NEO the past several years?

I’ve actually been able to find R-squared correlates up to the 85% mark, which indicates a level of statistical significance. I’m in the process of developing a model that can better forecast the price of NEO. It is a work in progress, but it’s definitely interesting. I would welcome you guys to take a look. I’ve posted my work and findings thus far in an article on my Github, while trying to explain things in layman’s terms. You can see it here:

Take a look and let me know your thoughts.

Best Regards,



im interted on NEO forecasr model but i have limited knowledge on R
Also will be very interesting to see the difference in correlation before and after N3 hard fork.

1 Like

If I were working on this project, I would investigate the following features first to build a black box model:

  • Total market capitalization of the sector (as well as first and second derivative)
  • Total market capitalization (excluding bitcoin, USDC, and USDT) (as well as first and second derivative)
  • Unique addresses and primary statistics for this distribution (median, mean, shape etc…)
  • Social Media(twitter, telegram, reddit) mentions and sentiment (first and second derivative)

This will also require some tuning for a good time-window since your derived features are going to be really noisy. You could try starting at 1 month of historical data for generating those features.

1 Like