Food Frequency Questionnaire Personalisation Using Multi-Target Regression.

Area of Science:

• Digital Health
• Machine Learning Applications
• Nutritional Science

Background:

• Mobile health applications aim to promote lifestyle changes for improved well-being.
• A nutrition self-monitoring module using a Food Frequency Questionnaire (FFQ) was developed for a health app.
• The 24-question Mediterranean diet FFQ, while informative, can be overwhelming for users.

Purpose of the Study:

• To reduce the number of questions in a dietary assessment questionnaire using machine learning.
• To enhance user experience in a mobile health application by optimizing data collection.
• To investigate machine learning methods for efficient dietary habit analysis.

Main Methods:

• Developed a machine learning model that utilizes previous user answers for targeted question selection.
• Compared the proposed method against random question selection and feature selection techniques.
• Conducted experiments using a multi-target regression approach to predict multiple nutritional goals simultaneously.

Main Results:

• The proposed machine learning method significantly reduced the number of questions required for dietary assessment.
• The model demonstrated superior predictive accuracy compared to random and feature selection methods.
• The approach effectively identified user-specific nutritional goals needing attention with minimal error.

Conclusions:

• Machine learning offers an effective solution for optimizing dietary assessment in mobile health apps.
• Reducing questionnaire length through intelligent methods improves user engagement and data collection efficiency.
• This approach supports personalized health interventions by accurately analyzing dietary habits.