Skip to main content
. Author manuscript; available in PMC: 2022 Dec 19.
Published in final edited form as: Proc ACM Interact Mob Wearable Ubiquitous Technol. 2020 Mar 18;4(1):23. doi: 10.1145/3381001

Table 5.

The 23 features selected by sequential forwarding selection for support vector classifier. K indicates an iteration number in feature selection.

K Accumulated features F1
1 Audio amplitude (24H) 0.657
2 Audio amplitude (6PM-0AM) 0.655
3 Variability of time spent at “religious” location 0.642
4 Distance travelled (24H) 0.668
5 Variability of activity “walking” duration (6PM-0AM) 0.680
6 Number of place visited (6PM-0AM) 0.710
7 Variability of activity “still” duration (6PM-0AM) 0.775
8 Variability of number of place visited (24H) 0.698
9 Variability of audio amplitude (24H) 0.718
10 Number of unlock (0–9AM) 0.719
11 Regularity of time spent at “religious” location 0.727
12 Variability of audio amplitude (0–9AM) 0.719
13 Regularity of activity “running” duration (0–9AM) 0.746
14 Activity “still” duration (6PM-0AM) 0.743
15 Time spent at “religious” location 0.747
16 Variability of conversation duration (24H) 0.762
17 Distance travelled (0–9AM) 0.789
18 Variability of distance travelled (24H) 0.777
19 Activity “walking” duration (24H) 0.772
20 Regularity of time spent at “greek” location 0.759
21 Conversation duration (0–9AM) 0.766
22 Variability of number of unlock (6PM-0AM) 0.764
23 Regularity of number of unlock (6PM-0AM) 0.793