A Novel Hybrid Food Recommendation System Integrating Deep Learning

Abstract

Food recommender systems serve as a valuable tool in assisting users in making healthier dietary
choices. This paper introduces a hybrid food recommender system designed to address the limitations of
existing approaches, including the neglect of food ingredients, time considerations, the cold -start
problem for both users and food items, and community-driven recommendations. The proposed system
operates in two key phases: a content-based recommendation phase utilizing Long Short-Term Memory
(LSTM) networks and a user-based recommendation phase employing a learning-based clustering
technique for users and food items. Additionally, a holistic approach is integrated to incorporate timeawareness
and community-related aspects, ultimately enhancing recommendation quality and user
satisfaction. The integration of these advanced techniques results in a highly adaptive, contextually
relevant, and personalized food recommendation system that not only improves user engagement and
satisfaction but also encourages healthier eating habits