IoT-Powered BGCNet: Thyroid Disease Detection Enhancement in Healthcare with IBESA based Feature Selection

Early diagnosis and treatment of thyroid problems, ranging from hyperthyroidism to thyroid cancer, depend heavily on the identification of thyroid illness. Utilising an actual dataset from the University of California, Irvine (UCI), this work offers a thorough method for identifying thyroid disease (TD). Preparing the data is essential to guaranteeing the dataset's quality. The dataset is improved to make it more appropriate for deep learning (DL) applications by managing outliers, and normalizing data. The Improved Bald Eagle Search Algorithm (IBESA), a cutting-edge optimization method inspired by bald eagle hunting behaviour, is used in the feature selection phase. Reducing dimensionality without reducing the critical information required for a precise diagnosis, IBESA effectively finds the most relevant features in the dataset. This reduces computational expenses while simultaneously improving model performance. Bidirectional Gated Recurrent Units (Bi-GRU) and Convolutional Neural Networks (CNN) are integrated to classify thyroid disorders. The temporal dependencies in the data are modelled using Bi-GRU-CNN. The study's findings show that an integrated strategy has the potential to greatly increase the accuracy to 98.11% and efficacy of TD identification compared to other existing models.