Diabetes Prediction using Optimized Bi-GRU Model with Hybrid Feature Selection and Metaheuristic Optimization

Diabetes prediction is crucial for early diagnosis and treatment. This study employs a Bidirectional Gated Recurrent Unit (Bi-GRU) to introduce a methodology grounded in deep learning principles model optimized with metaheuristic algorithms. This paper focuses on analyzing a balanced synthetic diabetes dataset to predict the likelihood of diabetes in patients. The dataset undergoes a pre-processing phase wherein it is segmented into distinct features and target variables following its importation from a CSV file. By means of fitness assessment, Grey Wolf Optimization (GWO) and Particle Swarm Optimization (PSO) are employed to ascertain the optimal hyper parameters. By facilitating an effective approach for recognizing individuals predisposed to diabetes, this initiative aims to enhance medical research and clinical decisionmaking, ultimately fostering early detection and superior healthcare results. Subsequently, Deep Learning algorithms could be developed utilizing the acquired data to predict the onset of diabetes. The ultimate goal of this program is to assist in early diabetes detection, which can lead to timely medical intervention and improved patient outcomes. This suggests that GWO serves as a more potent optimization strategy for the diabetes diagnostic model.