Perbandingan Metode Naïve Bayes dan K-NN dengan Ekstraksi Fitur GLCM pada Klasifikasi Daun Herbal

A. Nurjulianty (Universitas Muslim Indonesia, Makassar)
Purnawansyah Purnawansyah (Universitas Muslim Indonesia, Makassar)
Herdianti Darwis (Universitas Muslim Indonesia, Makassar)

JURNAL MEDIA INFORMATIKA BUDIDARMA
Vol 7, No 4 (2023): Oktober 2023

Indonesia is a country with various types of herbal plants that have the potential to be very effective medicines. Herbal plants have been used since ancient times as natural medicines. One part that has health benefits is the leaves, however, there are many similarities between the different types of leaves. This research aims  to classify digital images of herbal leaves implementing the Naïve Bayes and K-Nearest Neighbor (KNN) methods with Gray Level Co-occurrence Matrix (GLCM) feature extraction. The dataset consisted of sauropus androgynus and moringa leaves with data collection in bright and dark scenarios. A total of 480 data which was divided into two parts, namely 80% for training data and 20% for testing images. The KNN distances used for comparison are Euclidean, Manhattan, Chebyshev, Minkowski, and Hamming. Meanwhile, Naïve Bayes uses Gaussian, Multinomial, and Bernoulli kernels. The results of the study showed that the KNN method with the Manhattan distance obtained the best results with an accuracy rate of up to 94% in bright scenarios.