Mobile Information Systems / 2016 / Article / Tab 12

Research Article

A Mobile Application for Easy Design and Testing of Algorithms to Monitor Physical Activity in the Workplace

Table 12

Final bagging active/not active classifier performance, also considering the same classifier with a different smartphone position (upside down).
(a) Performance by cumulative indexes

Correctly classified instances 98.82% (98.83%)
Incorrectly classified instances 1.18% (1.17%)
Kappa statistic 0.969 (0.968)
Mean absolute error 0.019 (0.023)
Root mean squared error 0.09 (0.1)
Relative absolute error 5.26% (6.17%)
Root relative squared error 21.94% (23.18%)

(b) Confusion matrix

Classified as
ActiveNot active

Active 13853 (13825) 83 (111)
Not active 134 (105) 4339 (4368)

(c) Detailed accuracy by class

Precision Recall-scoreClass

0.99 (0.992) 0.994 (0.992) 0.992 (0.992) Active
0.981 (0.975) 0.97 (0.977) 0.976 (0.976) Not active

0.988 (0.988) 0.988 (0.988) 0.988 (0.988) Weighted average