Nomics Machine-Learning Tool Offers 7-Day Price Forecast on Top 100 Cryptos

Data provider Nomics is using machine learning to predict the future prices of cryptocurrencies like bitcoin.

Launched Thursday, the 7-Day Asset Price Prediction feed will give an outlook on future crypto prices based on purpose-built algorithms and the firm’s API, Nomics CEO Clay Collins told CoinDesk in an interview.

“There are a lot of poor signals out there that are getting a lot of clicks and we thought we could do a net positive for the space by just leveling up the quality of predictions,” Collins said.

The Nomics forecaster isn’t a standalone, investment-grade product, Collins added, but can help inform crypto investors based on curated exchange data.

The free tool currently lists 100 of the top cryptocurrencies by market cap. Collins said assets with longer histories and better data sources tend to lead to better predictions overall.

A subset of artificial intelligence, machine learning uses computer algorithms combined with data inputs to predict future events. Interestingly, machine learning services should get better given how these systems learn from past data and predictions.

Under the hood, the tool inputs “aggregate pricing data” from major exchanges that Nomics scrapes and cleans. That data is then fed into a purpose-built algorithm to spit out seven-day predictions in both a spot price and percentage change from the current price.

The feed, which has been in the works for the past six months, Collins said, relies on the recurrent neural network (RNN) architecture known as long short-term memory (LSTM). 

“LSTMs are relatively new in the machine-learning space, and financial data is notoriously difficult to predict, but we were able to get fairly reasonable predictions, and we are fully transparent about their historical accuracy,” Nomics CTO Nick Gauthier said in a statement.

Read more: Securitize Users Can Now Directly Trade Security Tokens With Just a Web Link

Of course, price predictions in any market, particularly crypto, can be exceedingly difficult to pin down.

Collins said the team wanted to add a touch of professionalism with its data-guided predictions. In the spirit of transparency, Nomics will also include a 30-day mean error rate with the new feature.

“We don't know when weird things are going to happen,” Collins said. “If the Yale endowment said they were going to allocate 30 percent in bitcoin – well, our models are not going to know that.”