Skip to main content
. 2019 Jun 7;2(1):e12153. doi: 10.2196/12153

Table 4.

Three-class classification results for proportional odds models, decision tree, and random forest on Wearable, Resident Assessment Instrument-Home Care, and Wearable+Resident Assessment Instrument-Home Care.

Dataset and classifier Accuracy Recall Precision F1
Wearable




POMa 0.378 (+/-0.198) 0.351 (+/-0.177) 0.267 (+/-0.147) 0.286 (+/-0.151)

DTb 0.595 (+/-0.184) 0.559 (+/-0.156) 0.389 (+/-0.186) 0.443 (+/-0.184)

RFc 0.649 (+/-0.166) 0.622 (+/-0.215) 0.550 (+/-0.242) 0.568 (+/-0.223)
RAI-HCd




POM 0.568 (+/-0.211) 0.509 (+/-0.216) 0.417 (+/-0.238) 0.449 (+/-0.229)

DT 0.703 (+/-0.218) 0.649 (+/-0.232) 0.554 (+/-0.299) 0.581 (+/-0.270)

RF 0.703 (+/-0.288) 0.662 (+/-0.299) 0.649 (+/-0.321) 0.634 (+/-0.314)
Wearable + RAI-HC




POM 0.676 (+/-0.170) 0.626 (+/-0.195) 0.593 (+/-0.195) 0.584 (+/-0.191)

DT 0.757 (+/-0.221) 0.703 (+/-0.254) 0.643 (+/-0.275) 0.662 (+/-0.266)

RF 0.838 (+/-0.199) 0.775 (+/-0.233) 0.730 (+/-0.259) 0.748 (+/-0.248)

aPOM: proportional odds model.

bDT: decision tree.

cRF: random forest.

dRAI-HC: Resident Assessment Instrument-Home Care.