Intelligent Fusion of Heuristically Optimized 1DCNN with Weighted Optimized DNN for Thyroid Disorder Prediction Framework

The early stage of thyroid disease prediction is useful to decrease the mortality and morbidity rate and also to increase the diagnosis efficiency of the patient-specific treatments. Existing thyroid prediction approaches suffer from several limitations because of unreliable human false-positive predictive outcomes. The deep learning-based diagnosis methodology provides higher prediction accuracy and earlier detection of thyroid disorders from the collected data set. However, the researchers face several challenges in the prediction of thyroid nodules from large dimensional datasets with higher prediction accuracy. Hence, this research aims to implement an efficient and Hybrid Deep Learning (HDL) for thyroid prediction to provide better treatment for thyroid disorder. Initially, the experimental data are obtained from standard datasets. The collected data are normalized using the data normalization technique. The normalized data are further utilized in the optimal feature selection, which is carried out using the Hybrid Artificial Gorilla Troops Sandpiper Optimization (HAGTSO) to get the optimally required features for thyroid prediction. The thyroid disorder can be identified using the HDL with One-Dimensional Convolutional Neural Network Model (1DCNN) and Deep Neural Network (DNN), where parameters in 1DCNN and weight in DNN get optimized using the developed HAGTSO. The experimental results demonstrate that higher performance is provided by the newly developed thyroid predictive model when compared to other comparative algorithms while considering the negative and positive metrics. The numerical analysis of the offered model shows 97% and 96% in terms of accuracy and specificity measures. Here, the designed model proved that it shows better performance than the existing methods.