Machine Learning Models for Classifying Physical Activity in Free-Living Preschool Children

Abstract

Machine learning (ML) activity classification models trained on laboratory-based activity trials exhibit low accuracy under free-living conditions. Training new models on free-living accelerometer data, reducing the number of prediction windows comprised of multiple activity types by using shorter windows, including temporal features such as standard deviation in lag and lead windows, and using multiple sensors may improve the classification accuracy under free-living conditions. The objective of this study was to evaluate the accuracy of Random Forest (RF) activity classification models for preschool-aged children trained on free-living accelerometer data. Thirty-one children (mean age = 4.0 ± 0.9 years) completed a 20 min free-play session while wearing an accelerometer on their right hip and non-dominant wrist. Video-based direct observation was used to categorize the children’s movement behaviors into five activity classes. The models were trained using prediction windows of 1, 5, 10, and 15 s, with and without temporal features. The models were evaluated using leave-one-subject-out-cross-validation. The F-scores improved as the window size increased from 1 to 15 s (62.6%–86.4%), with only minimal improvements beyond the 10 s windows. The inclusion of temporal features increased the accuracy, mainly for the wrist classification models, by an average of 6.2 percentage points. The hip and combined hip and wrist classification models provided comparable accuracy; however, both the models outperformed the models trained on wrist data by 7.9 to 8.2 percentage points. RF activity classification models trained with free-living accelerometer data provide accurate recognition of young children’s movement behaviors under real-world conditions.

Psychometric properties of instruments to measure parenting practices and children’s movement behaviors in low-income families from Brazil

“…Burdette et al [ 27 ] used the average vector magnitude recorded by the RT3 tri-axial accelerometer as a criterion measure of physical activity which serves as an indicator of the total volume of physical activity and not the type and intensity of physical activity in which children were participating. In the current study, a validated, state-of-the-art machine learning physical activity classification model for free living pre-schoolers was applied to the accelerometer data to derive physical activity metrics that captured both overall movement and time in energetic play [ 30 ]. Therefore, the criterion physical activity measure used in the current study may have had less random measurement error, resulting in stronger associations between self-reported and device-based measures of children’s physical activity behaviour.…”

“…Daily time spent in physically active movement behaviors was measured using the ActiGraph GT3X+ accelerometer (ActiGraph Corporation, Pensacola FL, USA). Raw accelerometer data (30 Hz) was downloaded and processed into physical activity metrics using a random forest physical activity classification algorithm specifically developed for children under five [ 30 ]. This validated machine learning algorithm uses 20 features extracted from the raw tri-axial acceleration signal to classify activity type and quantify daily time spent in sedentary activities (sitting or lying down), light-intensity activities and games (slow walking, standing, standing arts and crafts), walking, running, and moderate-to-vigorous intensity activities and games (active games with balls, riding bikes/scooters).…”

Psychometric Properties of Instruments to Measure Parenting Practices and Children's Movement Behaviors in Low-income Families From Brazil

“…Burdette et al [24] used the average vector magnitude recorded by the RT3 tri-axial accelerometer as a criterion measure of physical activity which serves as an indicator of the total volume of physical activity and not the type and intensity of physical activity in which children were participating. In the current study, a validated, state-of-the-art machine learning physical activity classi cation model for free living pre-schoolers was applied to the accelerometer data to derive physical activity metrics that captured both overall movement and time in energetic play [27]. Therefore, the criterion physical activity measure used in the current study may have had less random measurement error, resulting in stronger associations between self-reported and device-based measures of children's physical activity behaviour.…”

“…Daily time spent in physically active movement behaviors was measured using the ActiGraph GT3X + accelerometer (ActiGraph Corporation, Pensacola FL, USA). Raw accelerometer data (30 Hz) was downloaded and processed into physical activity metrics using a random forest physical activity classi cation algorithm speci cally developed for children under ve [27]. This validated machine learning algorithm uses 20 features extracted from the raw tri-axial acceleration signal to classify activity type and quantify daily time spent in sedentary activities (sitting or lying down), light-intensity activities and games (slow walking, standing, standing arts and crafts), walking, running, and moderate-to-vigorous intensity activities and games (active games with balls, riding bikes/scooters).…”

Age Differences in Estimating Physical Activity by Wrist Accelerometry Using Machine Learning

“…The proliferation of fitness trackers and wearable accelerometers offer an excellent opportunity to achieve this goal. The literature contains many examples of machine learning algorithms used for the processing and modeling of the accelerometer data including decision tree [ 8 ], random forest [ 8 , 9 ], bag-of-words [ 10 ], neural network [ 11 ] and others [ 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 ]. However, these models are often limited to a specific age group (e.g., adults 20–40 years old).…”