Performance Comparison of Hybrid CNN-XGBoost and CNN-LightGBM Methods in Pneumonia Detection

Pneumonia is an acute respiratory infection that affects the lungs. Pneumonia causes inflammation of the air sacs (alveoli) of the lungs, so the individual who has pneumonia experiences shortness of breath, coughing up phlegm, fever, and chills. Pneumonia can be detected through chest X-ray examination. A radiologist generally performs the acquisition of chest X-ray images. However, the availability of radiologists is limitation. Several machine learning applications have been used to classify chest X-rays automatically. The classification method used in this research is a hybrid Convolutional Neural Network (CNN)-Extreme Gradient Boosting (XGBoost) and Convolutional Neural Network (CNN)-Light Gradient Boosting (LightGBM). CNN is used as a primary classification method before proceeding to the main classification. The main classification methods for each classification method are XGBoost and LightGBM. Before classification, the chest X-ray image is processed through image preprocessing. Image preprocessing applied in this research includes grayscaling, image resizing, image stretching, and thresholding. The results of testing for the CNN-XGBoost classification obtained an accuracy value of 97.60%. While the results of testing for the CNN-LightGBM classification produced an accuracy value of 97.45%.