Artificial Intelligence-based Approaches to Assess Dietary Patterns: A Systematic Review

Development of artificial intelligence has provided novel opportunities for research in the field of nutrition sciences. This study was carried out with the aim of reviewing and comprehensively assessing studies linked to the field of diet and food patterns, using artificial intelligence and machine learning algorithms.

All studies published until June 2023 were searched using PubMed Cochrane, EMBASE and SCOPUS databases with associated keywords. No time and language restrictions were used.

After a complete review of the articles, 31 relevant articles were selected that were consistent with the purpose of the present study. Various machine-learning methods have various accuracy in predicting food patterns. For example, intelligent neural network is further accurate in predicting healthy food index quintiles, while it is further accurate in decision tree meals. Another use of machine learning includes extracting food patterns and investigating their relationships with various diseases such as obesity, heart diseases, stroke, risks of death from cardiovascular diseases and cancers. Machine learning methods such as decision trees are able to provide models for predicting adherence to various diets such as the Mediterranean diet.

Various artificial intelligence methods can help better understand food patterns linked to chronic diseases. The most important algorithms in the study of food patterns are decision tree, random forest, K-means, K-nearest neighbor, regression methods, support vector machine and intelligent neural network. These methods can help better understand dietary patterns associated with chronic diseases by categorizing and finding hidden associations between the groups and foods. Further studies are needed to better understand these connections.