Neural Network-Based Algorithm for Adjusting Activity Targets to Sustain Exercise Engagement Among People Using...

Area of Science:

• Health Informatics
• Machine Learning
• Behavioral Science

Background:

• Lack of physical activity negatively impacts health.
• Activity trackers promote monitoring but suffer from waning adherence.
• Motivating users with personalized goals can improve adherence to fitness targets.

Purpose of the Study:

• To develop a machine learning model for predicting achievable weekly activity targets.
• To enhance user adherence and utility of activity trackers through realistic goal setting.
• To integrate user activity patterns, behavioral, and environmental data for personalized goal prediction.

Main Methods:

• A neural network model was developed to prescribe realistic weekly activity targets.
• Inputs included personal, social, environmental factors, and 7-day step count entropy.
• Data from 20 participants over 9 weeks were used for training and evaluation.

Main Results:

• The model predicted target daily step counts with a mean absolute error of 1545 steps.
• The prediction accuracy was evaluated over an 8-week period.
• Data from 20 out of 30 enrolled participants were utilized.

Conclusions:

• AI-driven personalized goal setting using physical activity and behavioral data can improve fitness tracker adherence.
• This approach has the potential to increase user engagement with activity trackers.
• A prospective study is underway to assess the engagement algorithm's performance.