Thyroid Detection Model Using Support Vector Machine and Decision Tree

Abstract

As in the past few years’ technology is growing rapidly and healthcare is particularly one of them. Different methods have been proposed to predict the current nature of thyroid which includes samples through blood, fine needle aspiration. Biosensors are also used for thyroid detection, if some abnormal pattern occur Biosensor fabricated with MEMS (Micro-Electro-Mechanical Systems) analyze the Thyroid Stimulating Hormones (TSH) level through protein immobilized between the integrated electrodes (ELISA) and check the sensitivity of thyroid. The types of thyroid include Hyperthyroid, Hypothyroid and Normal thyroid. Therefore, to overcome the aforementioned constraints, the proposed scheme consists of three categorizes (i) Data denoising and Feature Extraction with wavelet transform to measure the data and dimensionally reduce the unwanted pattern or noise by keeping accuracy of original dataset by using machine learning models (ii) Feature Selection for filtering redundant features from dataset (iii) Classification Algorithms to provide a comprehensive evaluation. Finally, we evaluate that ‘Normal thyroid’ dataset doesn’t give much accuracy as compared to ‘hypothyroid’ and ‘hyperthyroid’. As in ‘Hyperthyroid’ Support Vector Machine (Svm) gives 99.7% accuracy and Decision Tree gives 99.9% accuracy on ‘Hypothyroid’ which performs ways better than previous studies.