A comprehensive study on thyroid diagnosis by neural networks and swarm intelligence

Diagnosis of the thyroid function abnormalities may take much precious time of the patient. So, a computer aided diagnosis system can guide physicians in diagnosis and can save time of the patient. In this study, seven different types of neural networks were implemented in order to realize more robust and reliable networks on thyroid diagnosis. The particle swarm optimization and artificial bee colony algorithms are well known optimization algorithms and the migrating birds optimization algorithm is a recently introduced optimization algorithm based on swarm intelligence. Finally, the designed feed forward multilayer neural network was re-trained by using these metaheuristic algorithms. A dataset in UCI machine learning repository web site was used. The results show that our accuracy values outperform the similar studies.