Perbandingan K-Nearest Neighbors dan Gaussian Naive Bayes pada Region of Interest Citra Mammogram Menggunakan Gray Level Co-Occurrence Matrix

Pratiwi, Ika Ayu (2026) Perbandingan K-Nearest Neighbors dan Gaussian Naive Bayes pada Region of Interest Citra Mammogram Menggunakan Gray Level Co-Occurrence Matrix. Tugas Akhir (S1) - thesis, Universitas Bakrie.

	Text (00. Cover) 00. Cover.pdf - Accepted Version Download (350kB)
	Text (01. BAB I-III) 01. BAB I-III.pdf - Accepted Version Restricted to Registered users only Download (494kB) | Request a copy
	Text (02. BAB IV) 02. BAB IV.pdf - Accepted Version Restricted to Registered users only Download (472kB) | Request a copy
	Text (03. BAB V) 03. BAB V.pdf - Accepted Version Restricted to Registered users only Download (212kB) | Request a copy
	Text (04. BAB Daftar Pustaka) 04. BAB Daftar Pustaka.pdf - Accepted Version Download (118kB)
	Text (O5. Lampiran) 05. Lampiran.pdf - Accepted Version Restricted to Registered users only Download (634kB) | Request a copy

Abstract

Breast cancer is one of the diseases commonly found in women and can be detected through mammography examination. Mammogram images have complex texture characteristics; therefore, a feature extraction method is required to represent the texture information contained in the images. Gray Level Co-Occurrence Matrix (GLCM) is one of the texture feature extraction methods that can represent the spatial relationships between pixels. After feature extraction, selecting an appropriate classification method is necessary because different algorithms may produce different performance on the same feature representation. This study aims to compare the performance of K-Nearest Neighbors (KNN) and Gaussian Naive Bayes (GNB) using GLCM features for mammogram image classification. The dataset consists of 322 unique mammogram images from the Mammographic Image Analysis Society (MIAS), consisting of 207 negative-class images and 115 positive-class images. The preprocessing stages include grayscale conversion, Region of Interest (ROI) determination, Contrast Limited Adaptive Histogram Equalization (CLAHE), and grayscale quantization. For abnormal images, the ROI was determined based on the lesion coordinates provided in the dataset annotations, while normal images used a center-crop approach. Feature extraction was performed using GLCM with a pixel distance of 3, four directions of 0°, 45°, 90°, and 135°, and 64 quantization levels. Six texture features were used, namely contrast, dissimilarity, homogeneity, energy, correlation, and Angular Second Moment (ASM), resulting in 24 features for each image. KNN was evaluated using different k values of 3, 5, 7, 9, 11, 13, and 15. Based on the experimental results, k = 5 was selected as the final configuration because it produced the highest akurasi and F1 Macro, with values of 0.6677 and 0.5916, respectively. Both methods were evaluated using Stratified 5-Fold Cross Validation with an out-of-fold approach based on akurasi, presisi, recall, F1-score, F1 Macro, and confusion matrix. The evaluation results show that GLCM + KNN with k = 5 achieved an akurasi of 0.6677, positive-class presisi of 0.5588, positive-class recall of 0.3304, positive-class F1-score of 0.4153, and F1 Macro of 0.5916. Meanwhile, GLCM + GNB achieved an akurasi of 0.6739, positive-class presisi of 0.6562, positive-class recall of 0.1826, positive-class F1-score of 0.2857, and F1 Macro of 0.5372. These results indicate that GNB achieved higher performance in terms of akurasi and positive-class presisi, whereas KNN achieved higher performance in terms of positive-class recall, positive-class F1-score, and F1 Macro. Therefore, neither method consistently outperformed the other across all evaluation metrics, and the selection of the classification method depends on the prioritized evaluation metric and classification objective. Keywords: Mammogram Image, GLCM, K-Nearest Neighbors, Gaussian Naive Bayes, Image Classification.

Item Type:	Thesis (Tugas Akhir (S1) - )
Uncontrolled Keywords:	Mammogram Image, GLCM, K-Nearest Neighbors, Gaussian Naive Bayes, Image Classification.
Subjects:	Computer Science Computer Science > Image Processing Thesis > Thesis (S1)
Divisions:	Fakultas Teknik dan Ilmu Komputer > Program Studi Sistem Informasi
Depositing User:	Ika Ayu Pratiwi
Date Deposited:	05 Sep 2026 02:48
Last Modified:	05 Sep 2026 02:48
URI:	https://repository.bakrie.ac.id/id/eprint/14256

Actions (login required)

View Item