Linear Kernel and Polynomial Analysis in Recognizing Tuberculosis Image Using HOG Feature Extraction

Ira Farenda Sudirman (Universitas Prima Indonesia)
Winda Hartati Giawa (Universitas Prima Indonesia)
Intan Permatasari Sarumaha (Universitas Prima Indonesia)
Sukurman Ndraha (Universitas Prima Indonesia)
Insidini Fawwaz (Universitas Prima Indonesia)

Jurnal Mantik
Vol. 4 No. 3 (2020): November: Manajemen, Teknologi Informatika dan Komunikasi (Mantik)

Tuberculosis (TB) is an airborne disease caused by mycobacterium tuberculosis (MTB) which usually attacks the lungs which can cause severe coughing, fever and chest pain. The recognition of TB negative and positive TB x-ray image patterns in this study uses HOG feature extraction and the SVM method as a classification method by adding linear and polynomial kernel functions to the SVM method. This is because even though it is very good at solving classification problems, SVM can only be used on linear data, so that in order to be used on non-linear data, SVM must be modified using kernel functions. The results showed that the linear kernel was better at classifying the x-ray image of TB with an average accuracy of 79.50% while the polynomial kernel was 77.50%.