The Diagnosis of Thyroid Diseases Using Combinati on of Neural Networks through Hierarchical Method

Problems in thyroid gland are more common than in other glands of human body, and if they are not diagnosed early, thyroid storm or myxedema coma is likely to happen that might lead to death; therefore, on-time diagnosis of thyroid disorders (Hypothyroidism or hyperthyroidism) based on Laboratory and clinical tests is necessary. The main object of this research was to present a model based on data mining techniques that is capable of predicting thyroid diseases.
This study was a descriptive-analytic study and its database included 7200 independent records based on 21 risk factors derived from UCI data reference. From all records, 70% were used for training and 30% for testing. First, neural networks performance was reviewed in order to diagnose thyroid diseases, and then an algorithm for combination of neural networks through hierarchical method was presented.
After modeling and comparing the generated models and recording the results, accuracies of predicting thyroid disorders using neural network and hierarchical method were found to be 96.6% and 100% respectively.
Reducing misdiagnosis of thyroid diseases has always been one of the most important aims of researchers. Using methods based on data mining can decrease these errors. This study showed that using combination of neural networks through hierarchical method improves diagnosis accuracy.