OHESV: Optimal hybrid ensemble support vector model for detecting and recommendation of food for diabetic patients

• Published in: Multimedia tools and applications 2024-01, Vol.83 (27), p.68907-68930
• Main Authors: Rachitha, M. V., Ramakrishna, M.
• Format: Article
• Language: English
• Subjects: Algorithms; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Datasets; Diabetes; Diet; Food; Food products; Internet of medical things; Multimedia Information Systems; Personal health; Recommender systems; Special Purpose and Application-Based Systems
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-17954-7

Maintaining a healthy diet is essential, particularly for diabetic patients who are affected by severe defects. Nutrition therapy is vital to avoid the effects of diabetes and maintain better health. However, high sugar, fatty foods are significantly avoided by the patients to find alternate food from the same group. Despite this, there are still variations in diet that affect diabetic patients. Therefore, this paper proposes an effective food recommendation system to suggest suitable dietary plans for patients affected by diabetes. The personal health information of patients and several food products gathered from the Internet of Medical Things (IoMT) dataset are taken as input data for the proposed integrated recommendation system. To boost unbiased detection, the data noise present in the IoMT dataset is eliminated and regularized using preprocessing steps, including data cleaning, data reduction, data transformation, data enrichment, and data validation. From the preprocessed data, significant feature characteristics that contribute more to diabetes detection and food product recommendation are selected using the Wingsuit Flying Search (WFS) algorithm. These selected features are then classified using the proposed Hybrid Ensemble Support Vector Deep Residual (HESDR) approach. This approach accurately classifies diabetes and non-diabetes patients and suggests top nutrient-dense food products for patients with diabetes. The analytical results show that the proposed HESDR approach achieves a high accuracy rate of about 98.3% compared to other methods. In conclusion, this paper proposes an effective food recommendation system that accurately classifies diabetic patients and suggests suitable dietary plans based on their health conditions. The approach effectively eliminates data noise and selects significant features using the WFS algorithm. The proposed HESDR approach accurately classifies diabetic patients and recommends nutrient-dense food products for them, achieving high accuracy rates.