Ecological Validity of a Random Forest Activity Classifier for Wrist-Mounted Accelerometer Data

Trost S, Pavey T, Gomersall S, Clark B

4th ICAMPAM, 10 – 12 June 2015, Limerick; PS1.35

Introduction Wrist-worn accelerometers are convenient to wear and associated with greater compliance. Previous work has relied on choreographed activity trials to train and test classification models. Thus, validity in free living contexts remains an open question. The aims of this study were to: 1) train and test a random forest activity classifier for wrist accelerometer data; and 2) determine if models trained on laboratory data perform well under free-living conditions. Methods 21 participants (mean age = 27.6 6.2) completed 7 lab-based activity trials (lying, sitting still, sitting active, standing still, standing active, walking, running) and a 24h free-living trial (N=16). Participants wore a GENEActiv monitor on the non-dominant wrist during all trials. Classification models to recognize four activity classes (sedentary, stationary plus, walking, and running) were trained using time and frequency domain features extracted from 10 sec non-overlapping windows. Model performance was evaluated using 10-fold cross-validation. Models were implemented using the randomForest package within R. Classifier accuracy during the 24h free living trial was evaluated by calculating agreement with concurrently worn ActivPAL monitors. Results Overall classification accuracy for the random forest algorithm was 97.6%. Recognition accuracy for sedentary, stationary plus, walking, and running was 96.6%, 97.7%, 97.9%, and 100%, respectively. Agreement with the ActivPAL data (stepping vs. non-stepping) during the 24h free living trial was excellent and exceeded 90% for participants. The ICC for stepping time was 0.92 (95% CI = 0.75 – 0.97). Mean bias was 10.3 min/d (95% LOA = -46.0 to %2B25.4 min/d). Discussion & Conclusion The random forest classifier for wrist accelerometer data yielded accurate group-level predictions under both controlled and free-living conditions. Future studies should conduct more rigorous field based evaluations using observation of video.

NA

Trost S, Pavey T, Gomersall S, Clark B Ecological Validity of a Random Forest Activity Classifier for Wrist-Mounted Accelerometer Data 4th ICAMPAM, 10 – 12 June 2015, Limerick; PS1.35 2015. http://ulir.ul.ie/handle/10344/4487.