An enhanced deep learning model with context-aware attention for diabetes prediction

Abstract

A plethora of people worldwide suffer from diabetes, a chronic, potentially fatal illness that resulted serious risks and complications if left untreated. Effective management requires early prediction and intervention. Despite their advantages, traditional machine learning techniques frequently find it difficultly in grasping the intricate temporal as well as geographical correlations included with in medical stats. For the purpose of effectively forecast diabetes mellitus, the proposed work suggests a unique deep learning model called multilayer diabetes deep learning attention with context mechanism (MLDDAM). This model incorporates a hybrid architecture that integrates an Attention with Context Mechanism to enhance the model’s efficiency will be conversant with emphasizing on key aspects, convolutional neural networks (CNN) are utilized to extract traits, and bidirectional long short-term memory (BiLSTM) captures sequential dependencies. This innovative design enables the model to perform better by utilizing the input data’s temporal and geographical properties. Experiments using benchmark datasets show that the suggested MLDDAM model is efficient and robust, with outstanding 99.43% prediction accuracy for diabetes. These outcomes demonstrate the MLDDAM model’s effectiveness as a precise and dependable tool to assist clinical decision-making in the management of diabetes.