Summary: A recent study analyzed data from nearly 12, 000 people to identify the vital predictors of physical activity commitment. According to the study, sitting day, gender, and education level were key indicators of whether someone adhered to regular exercise recommendations.
Researchers could more freely predict exercise habits by training models on data from lifestyle, demographic, and health survey data than they have with conventional methods. These reassurances and public health initiatives that are more efficient and tailored to specific needs may be influenced by these findings.
Important Information
- Major Predictors: Passive time, sex, and learning level were the most consistently reliable predictors of exercise adherence.
- Use of Machine Learning in the Study: Researchers used data from 11, 683 people who participated in a national health review.
- Possible Impact: The findings could influence personalized exercise regimens and influence health coverage.
University of Mississippi cause
Some people struggle to stick to a workout regimen. However, a study team at the University of Mississippi is discovering what keeps people motivated during exercise.
Seungbak Lee and Ju-Pil Choe, both PhD students in physical training, and Minsoo Kang, a professor of activity analysis at the Department of Health, Exercise Science and Recreation Management, are working together to determine whether someone is adhering to physical activity recommendations based on their body type, age, and populations.
They have analyzed the results of about 30 000 research. They’ve turned to machine learning, a method of using servers to make estimates and trends based on the information to quickly sort through such a large data collection.
The team’s findings, which were published in the Essence Investment journal Scientific Reports, are accurate, Kang said.
According to him,” Physical exercise compliance with the guidelines is a public health issue because of its connection to disease prevention and general health patterns.”
” We wanted to use sophisticated data analytical methods, such as machine learning, to model this behaviour,” said the company.
As part of a healthier life, the U.S. Department of Health and Human Services recommends that people aim for at least 150 minutes of moderate exercise per year, or 75 hours of strenuous exercise.
The average American spends only two days a week on physical exercise, according to research, half of the four days that the Centers for Disease Control and Prevention recommend.
Lee, Choe, and Kang used information from the government-sponsored National Health and Nutrition Examination Survey for the years 2009 to 2018.
According to Choe, the study’s lead author,” We aimed to utilize machine learning to determine whether people adhere to physical activity guidelines based on questionnaire data. We also sought to find the best combination of factors for correct projections.”
” Demographic variables like gender, age, race, education, marital status, and income were taken into account as well as demographic factors like waist circumference and BMI,” according to the report.
The researchers also took into account lifestyle variables like alcohol intake, smoking, work, sleeping patterns, and passive behaviour to understand how they affect a person’s physical activity, he said.
Even though each model identified different variables as important, the results consistently revealed that three key variables: how long someone spends sitting, their gender, and their education level, also showed up in all of the top-performing models for predicting exercise habits.
These factors are particularly crucial for understanding who is more active and socially connected, according to Choe, and they may guide future health recommendations.
” I anticipated that our prediction model would take into account factors like gender, BMI, race, and adolescence,” he said,” but I was surprised by how significant educational status was.” “Education status is an external factor, but gender, BMI, and age are more innate to the body.
The researchers excluded data from people who had previously had physical activity data during the analysis and who had other diseases. That extracted the pertinent information from 11, 683 participants.
The researchers claim that machine learning gives them more time to study the data. Older methods require things to follow a straight-line pattern, and they fail to function well when some information is too similar.
Since machine learning has greater flexibility, it can find patterns without those restrictions.
” Our study’s use of subjectively measured physical activity data, where participants recalled their activity from memory,” Choe said.
People frequently overestimate their physical activity when filling out questionnaires, so having more accurate, objective data would increase the study’s reliability.
The researchers claim that because of this, they could use a similar approach to research in this area in the future, including studying dietary supplements, using more machine learning algorithms, or relying on objective data rather than self-reported data.
That could aid in the creation of long-term workout regimens that people can stick to.
About this news about AI and exercise research
Author: Clara Turnage
Source: University of Mississippi
Contact: Clara Turnage – University of Mississippi
Image: The image is credited to Neuroscience News
Open access to original research.
Seungbak Lee et al.,” Machine learning modeling to predict fitness to the physical activity recommendation.” Scientific Reports
Abstract
Machine learning modeling to predict fitness to the physical activity recommendation
This study uses ML to create predictive models for PA guidelines and examine the crucial determinants that influence adherence to the PA guidelines. The National Health and Nutrition Examination Survey’s 11 and 638 entries were analyzed.
Variables were broken down into the categories of lifestyle, anthropometric, and demographic. 6 ML algorithms were used to create 18 prediction models, which were based on accuracy, F1 score, and area under the curve ( AUC) scores.
Additionally, we used permutation feature importance ( PFI ) to determine the variable significance in each model.
The most efficient method for predicting PA guidelines was the decision tree that used all variables ( accuracy = 0. 075, F1 score = 0. 819, and AUC = 0. 052 ) ( as demonstrated by the decision tree using all variables ).
According to the PFI, sedentary behavior, age, gender, and educational status were the most crucial factors.
These outcomes highlight the applications of data-driven research with ML in PA research.
Additionally, we identified important variables, which helped us identify potential strategies to improve people’s compliance with PA guidelines.
