New algorithm predicts depression more accurately than current methods.
Depression prediction from behavioral signals is best done using ordinal regression, which focuses on ranking. A new method called GREP helps design effective models for predicting depression, outperforming other methods like GSR and CNN.