An Efficient Early Prediction of Hypothyroid Disease with Ensemble Machine Learning Technique

Abstract

Physical illnesses, such as hypothyroidism, have recently become more prevalent. The topic is well-known in today's society. Hypothyroidism is a problem for the majority of people. The differences between the ratios of the normal and affected diagnostic reports act as a barometer for the illness. The condition of hypothyroid disease has already been the focus of various studies. We have found a few fantastic opportunities to further the technique. We advocate using effective algorithmic models to identify threats and spread early awareness. Our proposed method is uncomplicated to implement in the real world and suitable for simple hypothyroid illness forecasts. The dataset was housed on the Kaggle website. Different classifiers, including the Random Forest (RF), Logistic Regression (LR), Gradient Boosting (GB), K-Nearest Classifier (KNN), Adaboost Classifier (ABC), and Decision Tree (DT) methods, have been implemented in our model. Before feature selection, Random Forest (RF) given an accuracy of 97.09%, Logistic Regression (LR) given accuracy of 94.7%, Gradient Boosting (GB) given accuracy of 94.04%, Decision Tree given accuracy of 97.62%, Adaboost Classifier (ABC) given the accuracy 97.62%, K-Nearest Classifier (KNN) given the accuracy 95.1%. We have used ensemble techniques to get the best accuracy. Our voting classifier RDAGLK gave the best accuracy of 98.01%. After feature selection, Random Forest (RF) given an accuracy of 98.26%, Logistic Regression (LR) given accuracy of 95%, Gradient Boosting (GB) given accuracy of 94.7%, Decision Tree given accuracy of 97.27%, Adaboost Classifier (ABC) given the accuracy 94.55%, K-Nearest Classifier (KNN) given the accuracy 95%. We have used ensemble techniques to get the best accuracy. Our voting classifier RDAGLK gave the best accuracy of 99.09%.