AI trained on fundamentals says which stocks will crash in sync

Neural net models correlations without a returns history and could improve stress testing

An AI model trained only on slow-moving fundamental data has proven to equal traditional methods of determining which stocks might sink together in a sell-off.

The new approach allows investors to predict correlations for stocks with no returns history. It can also help in the creation of hypothetical scenarios for stress testing, quants say.

Charles-Albert Lehalle, professor at Ecole Polytechnique