AI-Driven Cloud Solutions for Scalable and Secure Diabetes Prediction in Healthcare Systems

Abstract

Diabetes is a fast growing global health problem affecting thousands across the world. Early diagnosis and effective management can help avoid the full blown consequences of complications such as heart disease and kidney failure. Conventional methods of diagnosis are often inaccurate, slow, and do not scale for larger datasets. This work proposes an AI cloud based diabetes predicting model that puts a premium on data security by using Blockchain along with AI/ML for decision making. The model aims to be secure, scalable, and efficient for predicting whether a patient is diabetic in terms of clinical parameters like glucose levels, BMI, age, and insulin. While the proposed system uses Blockchain technology for decentralized storage, feature extraction is conducted using Random Forest, and training is done using the Multilayer Perceptron (MLP). The proposed prediction model works on cloud infrastructure for training and deployment thereby ensuring scalability and real time access. The evaluation results indicated that the proposed solution surpassed the traditional methods in accuracy (98.25%), precision (98.46%), recall (98.46%), and AUC ROC (0.981). The proposed solution thus provides enhanced data security, scalability, and predictive accuracy and is an efficient and secure instrument for diabetes prediction for practitioners to intervene timely for better patient outcomes.