The aim of this study was to compare the energy expenditure (EE) estimates of activity-specific regressions (ASR), random forest (RFEE) and regression trees (treeEE) from raw accelerometer data in children. Forty-one children (age: 9.9 ± 2.2y) performed the activities of sitting, standing, walking, running, jumping, crawling, cycling and riding a scooter for 3.5 min., while 30 Hz raw accelerations were collected with one tri-axial hip-accelerometer and EE was measured using a portable device of spirometry. Twenty out of 42 accelerometer features calculated over 1-s windows were included into the prediction model of the RFEE according to their Gini-index. To provide the activity-specific information and the relevant features for the ASR, an a priori decision tree was used. The ASR accurately predicted the EE of sitting, walking, running, jumping, crawling and riding a scooter with biases of 0.04, 0.08, -0.33, -0.61, 0.08 and -0.41 MET, respectively. RFEE precisely estimated the EE of cycling, riding a scooter, jumping and running (bias: -0.18, -0.21, -0.57 and -0.29 MET) and the treeEE accurately predicted the EE of running and cycling (bias: -0.17 and -0.38 MET). The ASR predicted EE more accurately than RFEE or the treeEE. Using activity-specific information seems therefore to enhance the accuracy in assessing EE in children with raw accelerometer data.
© Copyright 2013 International Journal of Computer Science in Sport. Sciendo. All rights reserved.