Application of Machine Learning to Predict Dietary Lapses During Weight Loss

Area of Science:

• Behavioral science
• Machine learning
• Obesity research

Background:

• Dietary adherence is crucial for successful weight loss.
• Dietary lapses are predictable via physiological, environmental, and psychological indicators.
• Real-time prediction of lapses is feasible with technological advancements.

Purpose of the Study:

• To predict dietary lapses using machine learning techniques.
• To evaluate combining group and individual data for enhanced lapse prediction.

Main Methods:

• Trained and tested a machine learning algorithm for lapse prediction.
• Utilized ecological momentary assessment (EMA) via smartphone surveys.
• Collected data from adults with overweight/obesity in a 6-week weight control program.

Main Results:

• Group-level prediction accuracy reached 0.72 using WEKA decision trees.
• Generalization of group models to individuals was poor.
• Combining group and individual data improved prediction; 4 weeks of individual data recommended.

Conclusions:

• A predictive algorithm can enable in-the-moment interventions to prevent lapses.
• This approach can enhance weight loss outcomes.
• Methods are translatable to other health behavior lapse prediction.