Intelligent methods in managing of individual diet

• Дата: 2025
• Автори: Федорченко, Є. М., Олійник, А. О., Міхайлова, М. С., Зайко, Т. А., Степаненко, О. О., Федорченко, Ю. В., Федорончак, Т. В.
• Формат: Стаття
• Мова: Українська
• Опубліковано: Інститут проблем реєстрації інформації НАН України 2025
• Теми: greedy algorithm; linear programming; method Monte Carlo; genetic algorithm; Random Forest; XGBoost; neural measurement; individual diet; optimization; forecasting; BMI; basal metabolism; accuracy; fluidity; variability
• Онлайн доступ: https://drsp.ipri.kiev.ua/article/view/354582
• Назва журналу: Data Recording, Storage & Processing

The study examines an approach to generating personalized dietary recommendations based on a combination of optimization methods and machine learning techniques. The relevance of this research is determined by the need to develop intelligent decision-support systems in the field of healthy nutrition that are capable of considering individual physiological characteristics and dietary preferences of users. The proposed system generates individualized recommendations for the composition of a daily diet by taking into account key user parameters, including daily caloric requirements, body mass index, and basal To determine the optimal distribution of nutritional components, a comparative analysis of several optimization algorithms was conducted, including the greedy algorithm, the linear programming method, the Monte Carlo method, and the genetic algorithm. In addition, the effectiveness of forecasting the methods based on machine learning was investigated, particularly Random Forest, XGBoost, and a dense neural network. The use of such models makes it possible to account for complex relationships between user parameters, the caloric content of meals, and their nutritional value, which contributes to improving the accuracy of generating. The effectiveness of each of the considered approaches was evaluated according to several criteria, including the accuracy of the obtained results, computational speed, algorithm performance, and the variability of the generated diets. Experimental studies demonstrated that classical optimization methods show high accuracy but may have limitations in terms of solution diversity or computational efficiency. At the same time, machine learning algorithms provide better adaptation to individual user parameters and allow complex nonlinear relationships to be taken into account. The obtained results indicate that the combined use of optimization algorithms and predictive methods provides the best balance between accuracy, computational speed, and diversity of the proposed dietary options. The proposed approach can serve as a foundation for the development of intelligent diet planning systems focused on personalized user needs and the promotion of a healthy lifestyle. Tabl.: 4. Refs: 21 titles.