Photo credit: www.sciencedaily.com

Machine Learning Insights into Exercise Adherence

Maintaining a consistent exercise routine poses significant challenges for many individuals. To address this, a research team from the University of Mississippi is leveraging machine learning techniques to better understand the factors that influence workout commitment.

The team, comprising Seungbak Lee and Ju-Pil Choe, doctoral students focused on physical education, along with Minsoo Kang, a professor specializing in sport analytics within the Department of Health, Exercise Science and Recreation Management, aims to forecast whether individuals adhere to established physical activity guidelines based on their demographic data, body metrics, and lifestyle choices.

By analyzing approximately 30,000 survey responses, the researchers utilized machine learning algorithms to efficiently navigate through this extensive dataset. This approach allows for the identification of patterns and prediction of behaviors based on the analyzed information.

Publishing their findings in the journal Scientific Reports, part of Nature Portfolio, Kang highlighted the relevance of their work: “Adherence to physical activity guidelines is a critical public health issue given its correlation with disease prevention and overarching health trends. Our goal was to apply advanced analytical methods, such as machine learning, to anticipate this behavior.”

The U.S. Department of Health and Human Services’ Office of Disease Prevention and Health Promotion recommends that adults engage in at least 150 minutes of moderate exercise or 75 minutes of vigorous exercise weekly to support a healthy lifestyle. However, research indicates that the typical American achieves only about two hours of physical activity each week, which is far below the recommended four hours as outlined by the Centers for Disease Control and Prevention.

To carry out their study, Lee, Choe, and Kang utilized public data from the National Health and Nutrition Examination Survey, which spans from 2009 to 2018.

Choe, the lead author of the study, noted, “Our objective was to employ machine learning to ascertain whether individuals comply with physical activity guidelines based on survey data and to determine the most effective variable combinations for precise predictions. We factored in demographic elements like gender, age, race, education level, marital status, and income, alongside anthropometric data such as body mass index (BMI) and waist circumference.”

In addition, the team examined lifestyle factors like alcohol use, smoking habits, employment status, sleep quality, and sedentary behavior, seeking to understand their effects on physical activity levels.

The analysis identified three significant predictors of exercise habits: time spent sitting, gender, and educational attainment. These variables consistently appeared among the best-performing predictive models, revealing the complexity of exercise adherence.

Choe commented on the findings, observing that educational status emerged as a surprisingly significant factor, alongside more expected data points like gender and BMI. “While innate factors such as gender and age are inherent to individuals, educational status represents an external variable that profoundly influences physical activity,” he reflected.

In refining their dataset, the researchers excluded responses from individuals with specific health conditions and those who did not report physical activity data, narrowing their sample to 11,683 participants.

The researchers argue that machine learning offers a more flexible framework for examining data compared to traditional methods, which often assume linear relationships that may not capture the intricacies of human behavior. Machine learning excels in revealing patterns without these constraints.

Despite these advancements, Choe noted a limitation concerning the reliance on self-reported physical activity data, which can be prone to overestimation as participants rely on memory. “Future studies would benefit from employing more objective measures of activity, which would enhance the reliability of the findings,” he suggested.

In moving forward, the researchers are optimistic about applying similar methodologies to investigate additional factors, such as dietary supplement usage, and employing varied machine learning algorithms or more objective data collection techniques, to further enrich the understanding of exercise adherence.

Ultimately, these insights could inform trainers and fitness professionals, enabling them to design more effective and sustainable workout regimes that align with individuals’ lifestyles and capacities.

Source
www.sciencedaily.com