Multimodal hybrid reasoning methodology for personalized wellbeing services

Area of Science:

• Health Informatics
• Artificial Intelligence in Healthcare
• Personalized Medicine

Background:

• Traditional physical activity recommendation systems offer general advice, lacking personalization for individual needs and interests.
• Effective health promotion requires tailored strategies to encourage adoption of healthy habits and lifestyles.
• Personalized physical activity recommendations are crucial for improving user engagement and adherence to wellness goals.

Purpose of the Study:

• To propose a multimodal hybrid reasoning methodology (HRM) for generating personalized physical activity recommendations.
• To integrate rule-based reasoning (RBR), case-based reasoning (CBR), and preference-based reasoning (PBR) for enhanced personalization.
• To validate the HRM's effectiveness in a weight management scenario.

Main Methods:

• Developed a hybrid reasoning methodology (HRM) combining RBR, CBR, and PBR.
• RBR utilized physical activity guidelines, CBR leveraged expert experience, and PBR incorporated user preferences.
• Implemented and evaluated the HRM in a weight management context, comparing it against baseline and modified RBR systems.

Main Results:

• The proposed hybrid-CBR system demonstrated superior performance compared to baseline-RBR and modified-RBR systems.
• Achieved high performance metrics: 0.94% recall, 0.97% precision, and 0.95% f-score.
• The methodology effectively filtered irrelevant recommendations, minimizing Type I and Type II errors.

Conclusions:

• The multimodal hybrid reasoning methodology (HRM) significantly enhances the personalization of physical activity recommendations.
• This approach offers a more effective strategy for promoting healthier lifestyles and habit formation.
• The findings support the use of hybrid AI approaches in developing personalized wellness systems.