Comparative Analysis of Machine Learning Algorithms To Detect Pneumonia in Children

Abstract

Pneumonia is a lung disease caused by a viral or bacterial infection in the lungs. It is responsible for high child mortality and mor bidity rate all over the world. It is very important to detect it early for effective treatment and management of the disease. Traditional methods such as Chest X-Ray (CXR) and CT scans are used worldwide to detect pneumonia in children. However, due to advancements in ma- chine learning in recent years, machine learning models are being used to detect pneumonia in children using CXR. This study compares the different performance metrics such as accuracy, precision, F1 score, and efficiency of seven different machine learning algorithms like Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), K-Nearest neighbour (KNN), Gaussian Na ive Bayes (GNB), that analyze a dataset of 5,216 images of Chest X-Ray of children taken from Kaggle to identify patterns and features of the dataset and give us a prediction based on this model that whether a person has pneumonia or not. Experimental results showed that Random Forest per- formed the best with an Accuracy of 87.34%, Precision of 94%, Recall of 97% and F1 score of 91%