PERSON—Personalized Expert Recommendation System for Optimized Nutrition

Abstract

The rise of personalized diets is due to the emergence of nutrigenetics and genetic tests services. However, the recommendation system is far from mature to provide personalized food suggestion to consumers for daily usage. The main barrier of connecting genetic information to personalized diets is the complexity of data and the scalability of the applied systems. Aiming to cross such barriers and provide direct applications, a personalized expert recommendation system for optimized nutrition is introduced in this paper, which performs direct to consumer personalized grocery product filtering and recommendation. Deep learning neural network model is applied to achieve automatic product categorization. The ability of scaling with unknown new data is achieved through the generalized representation of word embedding. Furthermore, the categorized products are filtered with a model based on individual genetic data with associated phenotypic information and a case study with databases from three different sources is carried out to confirm the system.

Personalised nutrition and health

“…The Food4Me Study published algorithms to integrate information based on current diet, phenotypic characteristics, and genotypic characteristics. 18 However, other approaches—for example, using machine-learning 19 or artificial intelligence, 20 might offer additional advantages.…”

Application of Deep Learning in Food: A Review

“…It is not difficult to find that food identification, fruit, vegetable, fish, and meat quality detection tasks have begun to use deep learning technology, but only a very few of articles can be found that applied deep learning for food calorie estimation, food supply chain, and food contamination problems, and also for menu recognition (Lee, Chiu, & Chang, ), nutrition measurement (Pfisterer, Amelard, Chung, & Wong, ), food safety risk early warning (Geng, Shang, Han, & Zhong, ), personalized diet recommendation (C.H. Chen, Karvela, Sohbati, Shinawatra, & Toumazou, ), dining experience (Naritomi, Tanno, Ege, & Yanai, ), visual modeling of food (P.Y. Chen et al., ), and so on, which we did not describe in detail in the survey.…”

Algoritmo Inteligente para Otimizar a Geração de Recomendação Alimentar com Muitos Objetivos

“…Estão disponíveis na literatura alguns trabalhos com algoritmos evolucionários que foram aplicados no contexto de geração de recomendações alimentares com diferentes abordagens. Em [5] e [6] os autores optaram por usar a otimização mono-objetivo com um AG (algoritmo genético) clássico. Já em [7] e [8] a abordagem multiobjetivo foi aplicada.…”

“…Na prática, os resultados do algoritmo com esta abordagem de avaliação se mostraram promissores durante os testes, com o aumento do número de recomendações oferecidas. Uma das limitações apontadas por [5]é o uso de porções de tamanho fixo dos alimentos, acarretando em recomendações limitadas a estas porções. Para contornar este problema, foi adotada a estratégia de permitir duas ou mais porções de um mesmo alimento em uma solução e utilizar recomendações com quantidade variável de alimentos para aumentar o número possível de resultados.…”