Thyroid Disease Prediction using Random Forest Algorithm
V. Vishnu Priya, R. Subashini, S. Hari Priya
Abstract
Recent years have seen an increase in the incidence of thyroid conditions. The pituitary hormone is most crucial besides regulating respiration. Two of the least prevalent conditions resulting from abnormalities in the thyroid gland are hyperactivity and hypothermia. Every year, thyroid conditions involving hypothyroidism and hyperthyroidism are found in a large number of people. Both hypothyroidism and hyperthyroidism can be brought on by a deficiency in the thyroid hormones levothyroxine (T4) and triiodothyronine (T3), which are generated by the thyroid parotid. To properly care for the individual just at right time and prevent unnecessary deaths and medical expenses, a strategic thyroid ailment prediction is necessary. Since each thorough health history is known over time, it was possible to predict how each patient would respond to therapy in the future and decide whether to intensify or reduce it based on the pattern of biological indicators and other factors. It is highly desirable to improve healthcare procedures to identify and prevent thyroid problems utilizing cutting-edge technologies. Numerous techniques are recommended in the literature for diagnosing thyroid disease. Machine learning approaches are increasingly used to forecast thyroid diagnoses as a result of improvements in data processing and computation before they are fully formed. The thyroid illness is categorized by using the data from GitHub repository. These algorithms, the same as SVM, KNN, Decision Tree, and Naïve Bayes produced results with an accuracy of up to 90%. Whereas an existing technique like the Random Forest algorithm produced an accuracy of about 70%. In order to improve the accuracy, this study has introduced the dimensionality reduction technique like PCA, which was found to produce an accuracy of about 90%.