Thyroid gland is one of the largest and most important organs in the body, which is responsible for regulating the metabolism. Timely diagnosis of thyroid disease reduces the mortality rates. Traditional diagnosis methods depend on experience and expertise of radiologists and pathologists. Several methods have been proposed for disease diagnosis and prevention, deep learning-based models have provided more accurate results in this regard. However, there is still need for improvement. In this thesis, we employed data mining methods to diagnose and classify different thyroid diseases. The innovative aspect of this study is the combination of LSTM neural network and whale optimization algorithm. In fact, in this work, using whale optimization algorithm, the optimal values of learning rate and batch size of the neural network were determined using this meta-heuristic algorithm. The datasets utilized in this study were obtained from the UCI repository. Finally, using the proposed method, we were able to achieve 97.1% accuracy, which was superior to other state of the art methods.
