Enhanced Diabetes Prediction Using Deep Learning Framework: Enhanced Diabetes Classification with CNN and Bidirectional LSTM Hybrid Approach

One of the main global health problems is diabetes, which frequently results in serious conditions such kidney damage, cardiovascular illnesses, and vision loss. Reducing these effects and enhancing patients' health outcomes are largely dependent on early diabetes identification. A mixed deep learning architecture is utilized in the proposed study to accurately forecast diabetes. The model combines Bidirectional Long ShortTerm Memory (BiLSTM) networks for containing temporal and sequential dependencies with Convolutional Neural Networks (CNN) for identifying spatial patterns in medical data. This combination makes it possible for the system to detect time-related and spatial characteristics that are essential for predicting the onset of diabetes. Additionally, the most important features are chosen using Particle Swarm Optimization (PSO), an optimization method inspired by nature, which improves the model's accuracy and generalization. The proposed framework is calculated using standard metrics, representing its high predictive reliability. This study highlights the prospective of integrating sophisticated AI methods to build an efficient tool for early detection of diabetes, facilitating healthcare providers to formulate timely and informed decisions.