Pneumonia Classification Based on GLCM Features Extraction using K-Nearest Neighbor

Suharyana Suharyana (Physics Department, Universitas Sebelas Maret, Surakarta)
Fuad Anwar (Physics Department, Universitas Sebelas Maret, Surakarta)
Armylia Chandra Dewi (Physics Department, Universitas Sebelas Maret, Surakarta)
Mohtar Yunianto (Physics Department, Universitas Sebelas Maret, Surakarta)
Umi Salamah (Informatics Department, Universitas Sebelas Maret, Surakarta,)
Rifai Chai (Engineering Technologies, Swinburne University of Technology, Melbourne)

INDONESIAN JOURNAL OF APPLIED PHYSICS
Vol 13, No 2 (2023): IJAP Volume 13 ISSUE 02 YEAR 2023

Pneumonia has been detected using Machine learning. The stages in this study began with preprocessing in 4 stages: resizing, cropping, filtering using a high pass filter, and Adaptive Histogram Equalization. The feature extraction process continued with 22 Gray Level Co-occurrence Matrix (GLCM) features and classification using K-Nearest Neighbor (KNN). The image used was 150 data sets for training on the classification of 3 classes with a ratio of 50:50:50 while training on two classes was 50 bacterial pneumonia and 50 viral pneumonia. The most optimal training data accuracy results were obtained using the angle direction on the GLCM, namely 135o with the KNN classification (k = 3). For the classification of two classes Using 40 data sets, an accuracy of 91% was obtained, while testing for three classes with 60 data sets was 83.3%.