Particle Swarm Optimization for dietary recommendations

Recommender systems find relevant information that is correctly tailored to the consumers’ preferences. Diet recommender systems provide alternative food recommendations that not only satisfy the user’s interests but also adhere to the best dietary practices. This study introduces an adaptive meal recommendation system that provides consumers tailored, nutrientdense items that take into consideration both their preferences and needs. The proposed system uses a swarm intelligence based approach to generate alternate food recommendations for food that the user wishes to substitute in their diet, ensuring that their nutritional requirements are met and their behaviour in selecting food displays an upward trend. A recommendation is modelled as a particle in the swarm and the nutrient constituents are the dimensions. The binary particle swarm algorithm generates a set of solutions and over the iterations improves the global best solution which is the recommended alternative food item. The local and global acceleration coefficients are maintained as constants; the sigmoid function is used to update the position of each particle. The goodness of the recommendations of the system was evaluated by conducting surveys with clinicians and other users too. The recommendations of the system had 21.93 units higher fitness score compared to the recommendations of the clinicians. The system can be used by clinicians, dieticians, and also general users who seek alternative food for a food item.