Quants use AI to shush noisy order-book data

Signals from clusters of seemingly informed trading perform better, researchers say

A team of academics has used a simple machine learning algorithm to filter out so-called noisy trading in stock exchange data and say they have generated more-powerful versions of signals already popular with quants.

The researchers – from the University of Oxford, University of California Los Angeles, Queen Mary University of London and Memorial University of Newfoundland – used an unsupervised