Table 3 F1 Scores and micro-average AUC for multi-class classification tasks.

From: Trade-offs between machine learning and deep learning for mental illness detection on social media

Model

F1 score (95% CI)

Micro-average AUC (95% CI)

Support vector machine (SVM)

0.7593 (0.7505, 0.7680)

0.9570 (0.9545, 0.9597)

Logistic regression

0.7474 (0.7388, 0.7554)

0.9478 (0.9447, 0.9507)

Random forest

0.7454 (0.7369, 0.7544)

0.9434 (0.9399, 0.9464)

LightGBM

0.7767 (0.7687, 0.7847)

0.9566 (0.9534, 0.9594)

ALBERT

0.7917 (0.7842, 0.7992)

0.9676 (0.9655, 0.9698)

Gated recurrent units (GRU)

0.7765 (0.7685, 0.7849)

0.9569 (0.9537, 0.9599)

  1. 95% confidence intervals (CIs) were calculated using bootstrapping with 1000 samples
Back to article page