Deep learning and multitask learning no better than tree ensembles for student outcome prediction

All models achieved F1 scores between 0.574 and 0.583, with recurrent networks not outperforming gradient-boosted trees. Single-task and multi-task variants showed statistical equivalence in academic outcome prediction. Task correlation (r=0.589) did not translate into performance gains. Predictive power was primarily driven by cumulative and rolling engagement metrics.

Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.