Food recommendation using machine learning for physical activities in patients with type 1 diabetes

Abstract

Physical activities have a significant impact on blood glucose homeostasis of patients with type 1 diabetes. Regular physical exercise provides many proven health benefits and is recommended as part of a healthy lifestyle. However, one of the main side effects of physical activities is hypoglycemia (low blood glucose). Fear of hypoglycemia generally leads to the patients not participating in physical activities. This paper shows a proof of concept that machine learning can be used to create a personalized food recommendation system for patients with type 1 diabetes. Machine learning algorithms were designed to improve glycemic control and reduce the overcompensation of carbohydrate. First, a personalized model based on feedforward neural networks is developed to predict the blood glucose outcome during and after physical activities. Based on the personalized model and reinforcement learning, optimal food intakes will be recommended to the patient. Simulation results show that the proposed methodology has successfully maintained the blood glucose in the healthy range on a type 1 diabetes simulator during physical activities.