Table 3.
Classification result using selected features of phone data.
| Two classes or subclasses being predicted (number of data samples) | Features selected, n | Best MLa model | Average percent accuracy (SD) | Average percent recall (SD) |
| Steady (n=144) and Swing (n=234) | 4 | Random forest | 66.76 (4.94) | 80.93 (7.72) |
| Steady-remission (n=25) and Swing-drastic (n=75) | 36 | Random forest | 70.74 (6.62) | 77.58 (7.12) |
| Steady-remission (n=25) and Swing-moderate (n=159) | 7 | Random forest | 80.92 (5.34) | 95.50 (2.30) |
| Steady-depressed (n=119) and Swing-drastic (n=75) | 10 | Decision Tree | 66.18 (6.31) | 65.71 (6.99) |
| Steady-depressed (n=119) and Swing-moderate (n=159) | 34 | Random forest | 75.23 (3.75) | 88.99 (6.00) |
aML: machine learning.