Towards automatically generating meal plan based on genetic algorithm

With the rising of the concept of balanced diet, more and more people pay attention to the healthy meal plans. However, a meal plan is usually difficult to meet people’s taste preferences and health standards at the same time. Aiming at the problem, we propose a novel method to generate health meal plan for different users with meeting their taste. Specifically, we firstly calculate the recommendation score for each dish based on the ranking order of the recipe in the recommendation list, and the recommendation score is used to represent the user’s preference. The fat, carbohydrate, sodium and other nutrients in the recipe were then calculated based on the recipe’s ingredient content. The problem of meal plan generation is regarded as an optimization problem with constraints, that is, to maximize the recommended score of the meal plan under the condition that all nutrients in the meal plan meet the health standards. In this way, we approximately solve the problem of the balance between user’s taste and the recipe’s health by introducing a genetic algorithm. The experimental results show that the average recommended score of the meal plan generated by our method is 22.7, which indicates that the generated meal plan satisfy users’ preference, and a survey involved 81 volunteers also supported this conclusion. Besides, the experimental results show more than 80% of the meal plans generated by our method are healthy when comparing with the baselines.