Decreasing in misclassification of determination thyroid disease in Shoushtar town using tree boosting algorithm

Thyroid is a vital gland, which affect all of the body oragans such as heart, digestive system, kidney and so on. The intention of this research is to decreas in wrong determination of normal thyroid gland from abnormal using boosting algorithm. This algorithm is a powerful method in diagnosis and prognosis. It iteratively grows base classifer on a sequence of reweighted datasets then takes a linear combination of consequencs and we hope improves accuracy at final. A total of 103 patients’ data corrolated to November 2010 until November 2011 from Shoushtar salamat laboratory were analyzed for detemination thyroid gland state. Conventional decision trees and boosting decision trees were made for diagnosis normal thyroid gland from abnormal thyroid gland using R softwere vedersion 3.0.1. Our findings revealed that for conventional decision trees misclassification rate, sensitivity and specificity with test set were 0.088, 0.91 and 0.92 respectively. However these figures considered by boosting desion trees were 0.029, 0.955 and 1 crrespondingly. Conclution: The boosting decision trees had possibily superior sucsses in diagnosis normal tiroid gland ftom unnormal. So using boosting decisin trees propose in determination thyroid gland state.