<![CDATA[Malnutrition, both in the form of overweight and underweight, remains a global health challenge. Unhealthy urban lifestyles and limited access to appropriate nutritional interventions exacerbate this problem. Technology-based approaches such as machine learning and Large Language Models (LLM) offer opportunities to improve the effectiveness of dietary management. This study proposes the development of a machine learning-based and LLM-integrated diet program prediction and recommendation system applied to Cafe NUT Castle. The system was developed to digitize body composition data recording, predict diet programs (weight loss, weight gain, and body fat loss) using the Random Forest algorithm, and generate personalized initial diet recommendations through the integration of the Gemini Flash-Lite API. Based on the test results, the prediction model achieved an accuracy of 93% on the test data and 84% on 50 new datasets. Evaluation of the diet recommendations generated by LLM showed a feasibility level of 86.6% which was categorized as very feasible. These results indicate that the developed system is not only accurate in predicting diet programs but also effective in providing initial recommendations that can support decision-making in digital nutrition consultation services.]]>
Copyrights © 2025
