Table 3.
Performance of prediction at the post level and user level for the machine learning models. All numbers correspond to percentage values.
| Task | Post-level prediction (%), mean (SD) | User-level prediction (%),mean (SD) | ||||||||
|
|
M1 | M5 | M10 | M100 | M1 | M5 | M10 | M100 | ||
| Adult child | ||||||||||
|
|
Precision | 61.2 (0.04) | 63.2 (0.04) a | 61.6 (0.04) | 62.3 (0.03) | 81.4 (0.13) | 86.4 (0.17) | 85.7 (0.19) | 85.4 (0.19) | |
|
|
Recall | 81.8 (0.05) | 76.7 (0.06) | 78.7 (0.06) | 78.5 (0.04) | 94.3 (0.15) | 92.7 (0.17) | 93.4 (0.14) | 93.0 (0.18) | |
|
|
F1-score | 70.0 (0.04) | 69.3 (0.04) | 69.1 (0.04) | 69.5 (0.03) | 87.4 (0.09) | 89.5 (0.12) | 89.4 (0.12) | 89.1 (0.12) | |
| Spouse | ||||||||||
|
|
Precision | 83.7 (0.06) | 81.2 (0.03) | 81.6 (0.05) | 80.9 (0.05) | 81.9 (0.40) | 81.1 (0.33) | 82.4 (0.38) | 80.2 (0.52) | |
|
|
Recall | 57.7 (0.06) | 65.4 (0.60) | 64.9 (0.06) | 67.8 (0.07) | 87.5 (0.41) | 90.6 (0.34) | 91.4 (0.34) | 91.3 (0.33) | |
|
|
F1-score | 68.3 (0.05) | 72.5 (0.04) | 72.3 (0.05) | 73.8 (0.06) | 85.6 (0.26) | 85.6 (0.27) | 86.7 (0.25) | 85.4 (0.34) | |
| Other | ||||||||||
|
|
Precision | 14.8 (0.08) | 20.0 (0.10) | 21.6 (0.15) | 23.4 (0.12) | 66.6 (0.76) | 74.9 (0.54) | 77.6 (0.58) | 76.3 (0.46) | |
|
|
Recall | 20.9 (0.14) | 25.0 (0.16) | 21.7 (0.15) | 18.8 (0.12) | 24.8 (0.49) | 44.6 (0.78) | 42.6 (0.79) | 39.5 (0.68) | |
|
|
F1-score | 17.3 (0.10) | 22.2 (0.13) | 21.7 (0.15) | 20.9 (0.12) | 36.1 (0.60) | 55.9 (0.70) | 55.0 (0.75) | 52.0 (0.66) | |
| AUPRC | 71.6 (0.04) | 72.8 (0.04) | 72.8 (0.05) | 73.7 (0.05) | 75.9 (0.18) | 80.6 (0.25) | 81.3 (0.21) | 80.4 (0.26) | ||
aThe best performance for each metric is highlighted is italicized.