Using Machine Learning to Improve Physical Activity: A Systematic Review

BACKGROUND: Many diseases that threaten public health can be prevented with policies to improve physical activity. To improve physical activity, the implementation of individual interventions at the societal level remains limited. At this point, technological devices as mobile phones, which provide access to all individuals, enable widespread implementation of public health interventions and the cost-effective use of resources. In recent years, Machine Learning has been used in various areas of health promotion, including improving physical activity. This systematic review was conducted to examine the effect of machine learning on health outcomes related to physical activity. METHODS: Studies on this topic from 2013 to 2023 have been accessed by five database searches. The methodological quality of the studies was assessed using the Critical Appraisal Checklists for experimental and quasi-experimental studies, developed by the Joanna Briggs Institute. This study was conducted by following the Preferred Reporting System for Systematic Reviews and Meta-Analyses. RESULTS: This systematic review included thirteen studies with a total sample size of 2945 individual. In the study, it was found that using machine learning algorithms can help to detect physical actvivity level and intensity, increase the daily step count, predict adherence to the physical activity goal, develop physical activity skills, increase the duration of daily cycling time, accessing accurate and real-time data on physical activity, personalize physical activity recommendations and identify the types of gaits. CONCLUSIONS: Machine learning can be used to improve physical activity by providing more personalized and effective recommendations, using real-time feedback, and gamifying physical activity to increase motivation and adherence. Machine learning holds promise in the field of health promotion and has the potential to improve physical activity for individuals and communities. KEY MESSAGES: • Machine learning approaches are effective methods in determining and improving the level of physical activity on a public scale. • In terms of public health initiatives, using machine learning to improve physical activity is a cost-effective and efficient approach.