The effect of a machine learning-based mobile application on physical activity levels in overweight and obese women: A randomized controlled trial

Physical inactivity in women is a significant public health issue worldwide and in Turkey. Non-communicable diseases associated with physical inactivity, such as obesity and diabetes, are linked to high levels of mortality and morbidity. This thesis study aimed to evaluate the impact of a machine learning-based mobile application on the physical activity levels of overweight and obese women. The experimental study sample consisted of 80 women aged between 35 and 60 years with a Body Mass Index (BMI) of 25 or higher. Participants were randomly assigned to either an intervention group, which was provided with the machine learning-based mobile application, or a control group, which received standard physical activity recommendations. Physical activity levels were assessed using the International Physical Activity Questionnaire (IPAQ), daily exercise duration, and daily step counts. Data were monitored over a 12-week period, and cognitive and social factors affecting physical activity were collected on a monthly basis. The study found that, at the end of the third month, women in the intervention group showed a significantly higher increase in daily step counts, daily exercise duration, and IPAQ mean scores compared to those in the control group (p<0.05). Additionally, when evaluating factors influencing physical activity behavior, the total scores of the Cognitive Behavioral Physical Activity Scale, Exercise Self-Efficacy Scale, and Women's Physical Activity Self-Worth Scale were found to be significantly higher in the intervention group than in the control group at the three-month follow-up (p=0.000). The simplified XGBoost model, which incorporated physical activity data along with daily behavioral data, demonstrated the highest prediction performance (R²=0.50, RMSE=4.3) and identified self-efficacy, social and physical environmental factors, perceived barriers, daily step counts, and exercise feedback as the key determinants of adherence to physical activity goals. This thesis study demonstrates that the developed machine learning-based mobile application is effective in improving physical activity levels, cognitive-behavioral attitudes toward physical activity, exercise self-efficacy levels, and perceived self-worth related to physical activity in middle-aged women. It also highlights the importance of real-time data in personalizing interventions. Keywords: Physical Activity, Obesity, Machine Learning