Building a personalized diet proposal algorithm for covid 19 patients’ treatment at home with linear programming planning method

Abstract

Building a personalized menu for disease prevention and treatment is a complicated problem due to several reasons such as constraints on nutritional balance that vary for each specific user; the combination of type and amount of ingredients for each dish diverges greatly depending on personal preferences, the number of ingredients and dishes is excessively substantial, and so on. This article aims to build an algorithm providing an adequate diet for seven days based on personal nutritional information entered by the user. To achieve that goal, the algorithm needs to ensure three parts: the first is to determine the energy and volume of each nutritional ingredient suitable for the subject, the second is to determine the number of meals and the selection of dishes to eat pertinent to health requirements and the third is to determine the composition and volume of each dish so that the total nutritional value is close to the values calculated before. In part one, the "set of target nutritional indicators" is determined using nutritional calculation methods combined with the table of recommendations of the Ministry of Health and the Institute of Nutrition. The selected number of meals and dishes in part two and part three based on information from users and nutritional advice of nutritionists. 42 meals equivalent to about 70 dishes are randomly selected from a database to satisfy taste and nutrition requirements. Finally, the linear programming method is applied to solve the problem of determining the volume of ingredients in the dish so that the recommended menu and nutritional needs in 7 days are the lowest. In the linear programming, assuming that the amount of each course is a variable and the diet total nutrient content is the objective function that depends on those variables, the constraint from the difference of those with the "set of target nutritional indicators" to find the solution to optimize the objective function (i.e., minimize the difference). Compare that final result with the nutritional requirements from part one, repeat the selection step if the difference exceeds the allowable threshold mean that the selected dish is not suitable until the menu is appropriate. It is feasible to build a software or website recommending personalized diets for Covid-19 patients recovering at home based on this algorithm.