Automated recommender systems have been developed to make up for human inadequacies in decision making while solving the information overload problem. They have also found profound use in the area of diet and nutrition. Nevertheless, child nutrition in recommender systems is yet under researched, with very few works found in this area. This research work employs a switching hybrid recommendation technique that is a combination of user-based collaborative filtering and human expert knowledge for both healthy and malnourished children; to cater for the nutrition needs of children on a large and much improved scale while being accessible and available to children, parents and caregivers in different locations at the same time. Six elementary schools in Nigeria were visited for data gathering on children food interests, likes and dislikes. Open ended and dichotomous questions were used to obtain vital information for the system; and these responses are incorporated as initial user and food database to check the cold-start problem. Waterlows’ classification model was used to profile and classify users into their health classes and user-based collaborative filtering algorithm was used to recommend meals to the users based on user-user similarity. Human expert knowledge built from interaction with nutritionist was incorporated into the system and used in the recommendation process for both healthy and malnourished children. System evaluation results show the overall optimal performance and acceptance of the system. The results of this work can be adopted to reduce the scourge of malnutrition in children through healthy diet provisioning, especially in the Nigerian context.