Using decision-tree machine learning to identify worker movement typologies

Reasons for the study

Although physical activity has important health benefits, a minority of Canadians are regularly physically active. There are many barriers to physical activity, and these barriers are complex, operating at both individual and environmental levels. Work factors, such as job demands and social and physical working conditions, are also barriers to physical activity for many of the 18.5 million employed Canadians who spend most of their days at work.

Yet, many interventions to increase physical activity focus on specific factors. Few studies have explored the complexity of interrelated and, potentially, synergistic relationships. These include relationships among factors related to socioeconomics, health status, gender, social responsibilities, physical environment and work. This study aims to fill this gap.

Objectives of the study

  • Use Classifying and Regression Trees (CART), a form of supervised decision-tree machine learning, to determine worker characteristics associated with movement patterns at work and outside of work
  • Compare CART findings to those obtained through traditional, regression-based statistical methods
  • Understand whether supervised decision-tree machine learning offers a complimentary perspective to understanding complex worker characteristics (i.e. typologies) associated with movement patterns at work and outside of work
  • Build a ‘proof-of-concept’ for the use of decision-tree-based machine learning involving large-scale information (via data linkages) to identify typologies of work- and non-work-related physical activity
  • Understand how to more clearly collect information on upstream determinants of health related to correlates that affect individuals, working lives and society
  • Develop a baseline comparator for future, post-pandemic typology work classifications, allowing for assessments of how physical activity characteristics might have changed due to COVID-19 resulting in more work being done from home

Target audience

These findings will be of interest to public health experts, employers, workers and people generally who are interested in knowing how to measure movement patterns of workers at and outside of work and the relationship of these patterns to health.

Related scientific publications

Project status


Research team

Collaborators and partners

Memorial University
Public Health Agency of Canada