Development and Evaluation of a Food Recommendation System to Promote Healthy Eating

Abstract

Nowadays, food recommendation systems play a crucial role in promoting healthy eating behaviors. They assist individuals in making more informed dietary choices that align with their personal needs and preferences. Given the complexity of food selection, influenced by factors such as nutritional content, personal habits, and health goals, there is a growing need to develop intelligent systems capable of effectively supporting users in this decision-making process. Despite recent technological advances in recommendation systems, many traditional approaches still rely primarily on users’ past preferences and ratings, often overlooking the nutritional and health aspects of the recommended foods. This limitation hinders their ability to effectively support users in achieving their health-related objectives. To address this issue and strive for a balance between user preferences and health considerations through a fair and health-conscious recommendation strategy, we initially employed the NeuMF model to predict user preferences. In this approach, we integrated ingredient data and a health score to generate dietary recommendations that account not only for prediction accuracy but also for compliance with health standards. We also applied the NeuMF model a second time, this time excluding the health score and relying solely on ingredients. After this initial prediction, we used a linear function to combine the model’s output with the health score, aiming to produce recommendations that balance personal preferences with nutritional benefits. Striking this balance proved to be a significant challenge, as placing too much emphasis on health aspects can reduce user satisfaction and the overall effectiveness of the system