An Approach to Diagnosis of Thyroid using Data Mining Techniques

Abstract

The occurrence of thyroid as disease tends to affect the physical health of people across the world. The disease is expected to be present in the endocrine glands of a human body and appears to be in the anterior position with respect to the neck. An imbalanced production of hormones from such glands results into a dysfunction of thyroid glands due to insufficient generation of such hormones. Hence, the glands begin to swell and might lead to malignant tumours. One of the available treatments to for such a disease is being done using sodium levothyroxine; commonly referred to as LT4. It is known to be hormone for treating thyroid and its respective disorders. In addition to the treatments being made; it is necessary to predict and identify the occurrence of the disease in a human body so that it can be cured at the right time in the early stages. For this to occur; endocrinologists must detect the imbalance being created with respect to the generation of thyroid hormones. Hence, the proposed research work presents the detection and identification of the same by utilizing the concepts of data mining techniques. A database consisting of 857 patients is acquired from a repository and pre-processing stages are applied to it to refine and filter the collected data. Four data mining algorithms are further used to evaluate the accuracy thus produced and the algorithm with highest generating accuracy is thus declared as the optimised model. For the implementation of the proposed research paper; the execution of random forests generated highest accuracy of 81.26 percent.