Empowering Diabetic Care: Building an Edge-Enabled Machine Learning Model for Classification

Diabetes mellitus is a chronic metabolic dis-order characterised by elevated amounts of glucose in the bloodstream. The categorization and forecasting of diabetes are crucial for the effective administration of this prevalent long-term condition. Machine learning (ML) enables real-time monitoring and personalised therapy on edge devices. Machine learning can be employed to predict the likelihood of developing diabetes mellitus using advanced technology. Many symptoms are too subtle to be readily observed. It is a primary factor contributing to global blindness. It has a gradual lethal effect and there is currently no recognised remedy for it. This study explores the potential of edge-based predictive modelling for early identification of diabetes. Neural networks have the ability to accurately detect and analyse patterns and trends within patient data, leading to more accurate predictions of diabetes. Principal component analysis (PCA), quantization, dynamic pruning, and model compression are employed to optimise model performance while working within the limitations of the available re-sources. The proposed methodology is used and evaluated on the Pima Indian diabetic datasets, demonstrating its efficacy in achieving precise classification within the constraints of edge devices. Additionally, the possibility of shifting trained models from cloud servers to edge devices is being examined in order to reduce the computational burden.