A Novel Rule-Based Recommender System For The Indian Elderly Diabetic Population

Recommender systems are widely used for recommending items based on the user's specific preference. They depict user choices in a manner that can be exploited to personalize search results. In this paper, a novel rule-based model is proposed for recommending foods for Indian elderly diabetic population based on Glycemic Index (GI) of food items. Rules are extracted using Ripper algorithm from a real clinical dataset which is enhanced with the help of Synthetic Minority Oversampling Technique (SMOTE) and the food dataset used for this work is constructed from real data as the requirements of the proposed system are definitive. The proposed system is evaluated by medical professionals and doctors who rated the system based on a variety of use cases presented to them. The system received an average rating of 8 out of 10 on its performance to accurately identify the test results, GI range and appropriately suggest food items based on user preferences.