Table 8.

Ensemble vs. baseline performance (Threshold p = 80%, no. test samples = 4,071).

Model	Test Bal. Acc.	Test F1	Test Recall
Random Batching Ensemble	0.91	0.56	0.84
Subreddit Batching Ensemble	0.93	0.34	0.95
Baseline 1: Unbalanced LR	0.75	0.49	0.50
Baseline 2: Random Undersampling LR	0.90	0.33	0.91