An open-source tool to identify active travel from hip-worn accelerometer, GPS and GIS data

Abstract

BackgroundIncreases in physical activity through active travel have the potential to have large beneficial effects on populations, through both better health outcomes and reduced motorized traffic. However accurately identifying travel mode in large datasets is problematic. Here we provide an open source tool to quantify time spent stationary and in four travel modes(walking, cycling, train, motorised vehicle) from accelerometer measured physical activity data, combined with GPS and GIS data.MethodsThe Examining Neighbourhood Activities in Built Living Environments in London study evaluates the effect of the built environment on health behaviours, including physical activity. Participants wore accelerometers and GPS receivers on the hip for 7 days. We time-matched accelerometer and GPS, and then extracted data from the commutes of 326 adult participants, using stated commute times and modes, which were manually checked to confirm stated travel mode. This yielded examples of five travel modes: walking, cycling, motorised vehicle, train and stationary. We used this example data to train a gradient boosted tree, a form of supervised machine learning algorithm, on each data point (131,537 points), rather than on journeys. Accuracy during training was assessed using five-fold cross-validation. We also manually identified the travel behaviour of both 21 participants from ENABLE London (402,749 points), and 10 participants from a separate study (STAMP-2, 210,936 points), who were not included in the training data. We compared our predictions against this manual identification to further test accuracy and test generalisability.ResultsApplying the algorithm, we correctly identified travel mode 97.3% of the time in cross-validation (mean sensitivity 96.3%, mean active travel sensitivity 94.6%). We showed 96.0% agreement between manual identification and prediction of 21 individuals’ travel modes (mean sensitivity 92.3%, mean active travel sensitivity 84.9%) and 96.5% agreement between the STAMP-2 study and predictions (mean sensitivity 85.5%, mean active travel sensitivity 78.9%).ConclusionWe present a generalizable tool that identifies time spent stationary and time spent walking with very high precision, time spent in trains or vehicles with good precision, and time spent cycling with moderate precisionIn studies where both accelerometer and GPS data are available this tool complements analyses of physical activity, showing whether differences in PA may be explained by differences in travel mode. All code necessary to replicate, fit and predict to other datasets is provided to facilitate use by other researchers.Electronic supplementary materialThe online version of this article (10.1186/s12966-018-0724-y) contains supplementary material, which is available to authorized users.

The effect of moving to East Village, the former London 2012 Olympic and Paralympic Games Athletes' Village, on mode of travel (ENABLE London study, a natural experiment)

“…Another major strength was the use of an automated machine learning approach, combining accelerometry and GPS data, to measure travel mode allowing more data points to contribute to the analyses, increasing statistical power to establish the presence or absence of effects. The algorithm has been described previously and has major advantages over previously used manual approaches, which are prohibitively labour intensive particularly in larger studies [22].…”

“…Combined ActiGraph accelerometer and GPS travel recorder data were analysed using a previously described automated machine learning algorithm, which allocated each 10-s epoch of combined data to one of four travel modes, quantifying the daily time spent (i) walking, (ii) cycling, (iii) traveling by motorised vehicle (including car/van/bus/motorbike) or (iv) overground train. A fifth category classified recorded time where a journey was not taking place and the participant was stationary, e.g., sitting indoors at home or at work or stationary outside [22]. Gaps in the data due to loss of GPS signal were further classified as "underground" if the GPS signal was lost or regained within close proximity (200 m) of an underground station, and the time lapse between loss and regained signal was from 2 min to 2 h. However, as underground trains in the London transport system also run above ground, so there was potential for misclassification between "underground" and "overground train" modes of travel.…”

“…By combining data from accelerometers and global positioning system (GPS) monitors, machine learning tools are able to automatically discriminate between different travel modes [20,21]. For example, we have recently validated such an approach, making use of a gradient boosting machine learning tool [22]. There have been calls for high quality evidence to evaluate the effect of environmental interventions on health behaviours, particularly physical activity, making use of natural experiments where the population effects of change in travel infrastructure can be examined [23][24][25].…”

Socio-Ecological Natural Experiment with Randomized Controlled Trial to Promote Active Commuting to Work: Process Evaluation, Behavioral Impacts, and Changes in the Use and Quality of Walking and Cycling Paths

“…It should also be acknowledged that accelerometers were accompanied with an effortful diary approach in order to separate work trips from overall acceleration data. In recent studies, combination of accelerometer, GPS, and geographical information systems (GIS) data has been shown as a promising method to predict active travel and to decrease the workload for both participants and researchers [59,60].…”

Electrically assisted cycling for individuals with type 2 diabetes mellitus: protocol for a pilot randomized controlled trial

“…Raw Actigraph acceleration data will be extracted as csv files using ActiLife 6 software (Actigraph, FL, USA). Data from these devices will be merged by timestamp using an open-source tool, which will (1) classify different modes of transportation and (2) determine the amount of MVPA attributable to different active transport modes in the merged data [48]. This tool has been found to accurately identify active travel 94.6% of the time in a cross-validation study.…”