Effectiveness of a machine learning-based tailored Physical Activity program in Women with obesity

Abstract

Background Physical inactivity among women is a significant public health concern, contributing to increased rates of obesity. Tailored interventions leveraging technology may offer scalable solutions to enhance physical activity levels in this population. Methods A randomized controlled trial was conducted with 80 women aged 35 to 60 years and a BMI over 25. Participants were randomly assigned to an intervention group, receiving a tailored physical activity management program through machine learning-based mobile application, or a control group, receiving standard physical activity advice. The intervention lasted 12 weeks, with assessments at baseline and post-intervention. Primary outcomes included daily step counts, exercise duration, and International Physical Activity Questionnaire (IPAQ) scores. Secondary outcomes encompassed scores from the Cognitive Behavioral Physical Activity Scale, Exercise Self-Efficacy Scale, and Women Physical Activity Self-Worth Scale. Results At 12 weeks, the intervention group demonstrated significant improvements compared to the control group with increased daily step counts, longer exercise durations, and higher IPAQ scores (p < 0.05). Additionally, the intervention group showed greater enhancements in cognitive-behavioral determinants of physical activity, including self-efficacy and perceived self-worth (p = 0.000). The simplified XGBoost model, incorporating physical activity and behavioral data, achieved the highest predictive performance (R2=0.50, RMSE=4.3), identifying self-efficacy, perceived barriers, daily step counts, and exercise feedback as key predictors of adherence. Conclusions The machine learning-based mobile application effectively increased physical activity levels and improved cognitive-behavioral factors among women with obesity. Real-time data integration and personalized feedback were instrumental in enhancing adherence, highlighting the potential of digital health interventions in public health strategies. Key messages • A machine learning-based, tailored physical activity intervention significantly improved step count, exercise duration, and motivation in women with overweight and obesity. • Real-time behavioral data and cognitive-behavioral strategies enhanced adherence, highlighting the value of personalized digital tools in public health interventions.