Table 3.
Performance of several models predicting activity increase or decrease after edge creation using the area under the ROC curve. The table reports predictive performance on all data (all), and split by edge type and initiator: Friend (Fr), Follow (Fo), Sender (S), Receiver (R).
| Model | All | Fr/S | Fr/R | Fo/S | Fo/R |
|---|---|---|---|---|---|
| 1 None | 0.500 | 0.500 | 0.500 | 0.500 | 0.500 |
| 2 Previous behavior change | 0.538 | 0.543 | 0.551 | 0.546 | 0.526 |
| 3 Edge type, initiator and number | 0.574 | 0.515 | 0.518 | 0.506 | 0.510 |
| 4 User demographics (age, gender, BMI) | 0.685 | 0.644 | 0.583 | 0.777 | 0.773 |
| 5 Steps 7 days before | 0.715 | 0.665 | 0.614 | 0.808 | 0.781 |
| 6 All features (2–5) | 0.785 | 0.721 | 0.672 | 0.847 | 0.830 |