ADS Capstone Chronicles Revised

6

dietary recommendations through personalized and data-driven approaches. The review categorizes dietary recommendation systems into three categories: nutrition-based, recipe-based, and restaurant-based recommendations. These three categories help offer worthwhile insights that would be beneficial to the design of a diabetic-friendly recommendation system. Tsolakidis et al. (2024) explained the various methodologies used for personalized nutrition, highlighting how machine learning algorithms like collaborative filtering, content-based filtering, and hybrid models can help improve the accuracy of dietary recommendations. The review highlights that each model type has different strengths. For example, collaborative filtering learns from the preferences of similar users, content-based filtering focuses on nutrient profiles of foods, and hybrid models combine both approaches for an even more comprehensive recommendation (Tsolakidis et al., 2024). The methods were proved effective in predicting user preferences as well as their dietary needs, which makes this article highly relevant in the development of a diabetic-specific recommendation system (Tsolakidis et al., 2024). A diabetic-specific recommendation system would be heavily dependent on logic such as the hybrid models from Tsolakidis et al. (2024). The insights from Tsoloakidis et al. (2024) highlighted the significance of the potential of hybrid models, which could upgrade the tailored recommendation system by considering individual health data and dietary patterns. By using these machine learning models, the recommendation system can adjust to a user’s glucose management needs, creating a real-time solution. The approach aligns closely with the project’s goal of combining personalized health

data with dietary recommendations to ultimately create a healthier dining experience for diabetic patients.

3.4 A Personalized Restaurant Recommendation System Using ML-TOPSIS Approach

Previous studies explored the applications of recommendation systems to provide users suggestions on restaurants based on a myriad of factors. A Personalized Restaurant Recommendation System Using ML-TOPSIS Approach by Sulaiman et al. (2024) explored the use of machine learning tools to provide restaurant recommendations based on a user's geographical location, preferences, and health information including age, weight, height, and whether they have diabetes. Taking these factors into account, the two step system first creates a ranked list of restaurant recommendations using a multicriteria TOPSIS method based on calculated nutritional need, then collaborative filtering is deployed to refine the results and filter based on preferences and geographic location (Sulaiman et al., 2024). The calculated health fields in Sulaiman’s et al. (2024) study included body mass index, basal metabolic rate, and total daily energy expenditure which help the system determine generalized nutritional needs. While this approach is dynamic in the sense that it recognizes a user's updated restaurant preferences, it does not include a comprehensive method that uses dynamic nutritional needs. The methodology explored by Sulaiman’s et al. provides a foundation to build a system for more refined recommendations, involving the suggestion of specific meals at restaurants with the integration of real-time nutritional needs.

208