Deep Neural Network models for classification of significant attributes to predict Pre-Diabetes Mellitus

• Published in: International Conference on Power, Control and Embedded Systems (Online) pp. 1 - 4
• Main Authors: Singh, Balbir, Yadav, Jyoti, Singh, Manisha, Rani, Asha
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06.01.2023
• Subjects: Cross validation; Deep learning; Deep neural network; Early stopping; Machine learning; Neural networks; Particle measurements; Photoplethysmography; Physiology; Pre-Diabetes Mellitus; Predictive models; Wearable computers
• ISSN: 2643-864X
• DOI: 10.1109/ICPCES57104.2023.10076156

Hyperglycemia is a chronic condition associated with Pre-diabetes mellitus. It could lead to many health issues. According to recent increases in morbidity, the number of diabetic victims across the world is expected to reach 642 million by 2040, or one out of every ten persons. Without a doubt, this alarming number requires a lot of attention. However, detection of pre-diabetes condition will help the subjects to take precautions measures before head of actual diabetic condition. The Deep learning methods have shown its application in many facets of medical health. Therefore, Deep neural network (DNN) based system is proposed in the present work to detect pre diabetes using photoplethysmography (PPG) signal and other physiological signals. A dataset from 217 participants is used for experimentation. The obtained results are also compared with various machine learning (ML) methods. Further the results are also validated using 5-fold cross validation method. The result shows that DNN model performed well in the prediction of Pre-diabetes mellitus with 99.31% accuracy despite having fewer adjustable parameters. This obtained results shows that the proposed method can alert the subjects before onset of diabetes.