Deteksi Infeksi Mycoplasma Pneumoniae Pneumonia Menggunakan Komparasi Algoritma Klasifikasi Machine Learning

Abstract

Mycoplasma Pneumoniae Pneumonia is a pathogen that attacks the respiratory tract causing an infection. This infection generally occurs in school children and adolescents. Most of these infections are known as mild infections and go away on their own. However, there are cases of extrapulmonary manifestations including neurologic, dermatological, hematological and cardiac syndromes that can result in hospitalization and death. This can be minimized if early detection is carried out on people who are susceptible to the infection. One way is to apply machine learning. So that this can be achieved in this study, several machine learning algorithms will be used, namely Decision Tree, Logistic Regression, Gradient Boosting Decision Tree and Support Vector Machine. Each model will be modified on its hyperparameters using the Grid Search, Random Search and Hyperband methods. The final result shows that the hyperparameter modification method with Hyperband has a slightly better classification performance when compared to Grid Search and Random Search with f-score and accuracy values, namely 0.887 and 0.894 for Decision Tree, 0.942 and 0.947 for Logistic Regression, 0.910 and 0.915 for the Gradient Boosting Decision Tree and 0.591 and 0.715 for Support Vector Machine.