Recommendation Systems Used in the Field of Nutrition; A Systematic Review Study

Background and aims: Healthy diet is essential for health, physical and mental growth, performanceand productivity and overall well-being. Nutrition recommendation systems (NRS)facilitate changing eating behavior and making healthier food choices. NRS can also providerecommendations on diet and eating habits, detect health problems, and guide user behavior modification.The aim of this systematic review study is to investigate different recommendation systemsused in the field of nutrition.Method: This systematic review was conducted based on PRISMA guideline through an advancedsearch in PubMed, Scopus, Web of Science, IEEE (Institute of Electrical and ElectronicsEngineers) and GoogleScholar search engine without time limitation (up to February ۲۰۲۳). Thekeywords based on MeSH terms used include “Nutrition”, “Diet” and “Recommendation System”.After removing the duplicate titles using EndNote software, articles were screened andevaluated in three stages by reviewing the title, abstract and full text according to the inclusioncriteria (original articles, focusing on NRS and health promoting goals, papers with English languageand having access to full text). In order to be more accurate and lessening the error, screeningwas done by two individual reviewers in parallel and disagreements were discussed by thethird reviewer. Also review articles and letters were excluded. Related articles were examinedusing the MMAT tool. Finally, ۱۹ articles out of ۶۹۵ articles found in the initial search, were eligibleto be included in this study.Results: Studies show that hybrid recommender systems (۱۰ studies) and knowledge-based recommendersystems (۸ studies) were most used in NRS. Collaborative-based recommender systems(۴ studies) and content-based recommender systems (۳ studies) were also used in NRS.Frequently techniques used in NRS, were rule-based (۱۰ studies) and ontology (۶ studies). Othertechniques used are K-means (۶ studies), self-organizing map (SOM) (۳ studies), multi agents(۳ studies) and genetic algorithm (۳ studies). Mobile application was the frequented platformapplied in NRS (۷ studies). Web application (۵ studies) and desktop application (۱ study) wereother types of system’s platform.Conclusion: NRS have been designed in different types, with different artificial intelligence techniquesand different main modules on different system platforms. NRS has a significant potentialto improve dietary intake and promote a healthier life style if it is well designed, implemented andevaluated. Nutrition informatics specialists can design and develop a more useful NRS by beinginformed about the frequented types, AI techniques and modules.

کلیدواژه ها:

Nutrition ، Recommendation System ، Diet ، Informatics