RDED: Recommendation of Diet and Exercise for Diabetes Patients using Restricted Boltzmann Machine

Abstract

As per World Health Organization, noncommunicable diseases such as untimely birth, heart attacks, diabetes, and cancers are on the upswing. Diet intake that is insufficient or improper is known to cause a wide range of well-being illnesses. Due to the complexity of food components and a large number of dietary sources, it is difficult to select diets that must match one’s nutrition demands in real-time. Because of irrelevant material on proper food, patients are dependent on medicine rather than having precautionary steps in food consumption. Appropriate diet is especially crucial for persons living with chronic conditions and nutritionist food is essential for optimal health. An effective way to prevent disease is to eat a healthy nutritious diet. This study introduces the food and physical activity recommender system, which is capable of providing users with individualized and healthy nutrition recommendations based on their tastes as well as pathological medical data. Prescriptions characterize the ideal patient’s nutrition likes. In this paper, we show how Restricted-Boltzmann Machines, a type of two-layer undirected graphical model, can be utilized to describe ratings of food products. For this simple model, we provide effective learning and inference strategies that would be successfully applied to a food data set with over 100 million user-food ratings. When the predictions of the RBM model are created using different learning rates and several iterations, we attain an error rate of considerably below 0.30 percent in 50 epochs using 100 hidden nodes which fulfills our requirements. Hence, we want patients to use nutritious food rather than taking medicine to avoid an expensive trip to a physician.